floss final report – part 1 - · pdf filefloss final report – part 1 ... 6 survey...

Berlin, July 2002

Analyst:

Dr. Thorsten Wichmann

FLOSS Final Report – Part 1

Free/Libre Open Source Software: Survey and Study

Evidence from Germany, Sweden and UK Use of Open Source Software in Firms and Public Institutions

3

BERLECON RESEARCH GmbH

Oranienburger Str. 32

10117 Berlin

Tel.: +49 30 285296-0

Fax: +49 30 285296-29

Web: http://www.berlecon.de

Email: [email protected]

Acknowledgements:

This work was prepared by Berlecon Research. It is part of the final report for theproject „FLOSS – Free/Libre Open Source Software: Survey and Study“, which wasfinanced under the European Commission‘s IST programme, key action 4 as accom-panying measure (IST-2000-4.1.1).

Disclaimer:

The views expressed in this report are those of the authors and do not necessarily re-flect those of the European Commission. Neither the European Commission nor anyperson acting on behalf of the Commission is responsible for the use that might bemade of the following information.

Nothing in this report implies or expresses a warranty of any kind. Results from thisreport should only be used as guidelines as part of an overall strategy. For detailed ad-vice on corporate planning, business processes and management, technology integra-tion and legal or tax issues, the services of a professional should be obtained.

Names and trademarks mentioned in the report are the property of their respectiveowners.

© 2002 by Berlecon Research GmbH.

4

V 1.0 - 020630


5

Table of Contents

1 Introduction....................................................................................................... 9

2 Survey methodology ......................................................................................... 11

3 Professional use of Open Source software ......................................................... 153.1 OSS use in general....................................................................................... 163.2 OSS use by IT area...................................................................................... 17

4 Attitude to and importance of Open Source software in general ....................... 194.1 Attitude to Open Source software................................................................ 194.2 Importance of OSS within the IT infrastructure.......................................... 30

5 Benefits from using Open Source software ....................................................... 335.1 Benefits by usage area .................................................................................. 34

5.1.1 OSS used for server operating systems ................................................... 345.1.2 OSS used for databases .......................................................................... 375.1.3 OSS used on desktop computers ........................................................... 405.1.4 OSS used for creating and operating websites........................................ 445.1.5 Summary of OSS benefits for companies and public institutions........... 47

5.2 Establishment characteristics and OSS decisions.......................................... 48

6 Survey Questionnaire ....................................................................................... 55


6


7

Figures and tables

Table 2–1 Survey observations per strata .............................................................................. 13Table 2–2 Refusal rates in survey.......................................................................................... 13Figure 3–1 Usage of Open Source software by country.......................................................... 16Table 3–1 Current and planned professional use of OSS in Germany, Sweden and UK....... 17Table 3–2 Current and planned professional use of OSS in different application areas......... 18Figure 4–1 Independence from big software companies as motivation................................... 20Figure 4–2 Differences in importance of independence from big software companies............ 21Figure 4–3 Wish to support the OS community as motivation.............................................. 22Figure 4–4 Differences in importance of wish to support the OS community ....................... 22Figure 4–5 Availability of IT specialists as motivation ........................................................... 23Figure 4–6 Differences in importance of availability of IT specialists..................................... 24Figure 4–7 OSS use as company policy ................................................................................. 25Figure 4–8 Difference in relevance of OSS use as company policy......................................... 26Figure 4–9 OSS development during work time allowed....................................................... 27Figure 4–10 Differences in relevance of OSS development during work time.......................... 27Figure 4–11 Using OS service companies to support OS development.................................... 28Figure 4–12 Attitude to Open Source software in general ....................................................... 29Figure 4–13 Attitude to Open Source software in German establishments in general .............. 29Figure 4–14 Importance of OSS within IT infrastructure........................................................ 30Figure 4–15 Differences in importance of OSS within IT infrastructure ................................. 31Figure 5–1 Usage of different OS server operating systems .................................................... 35


8

Figure 5–2 Importance of all criteria for decision in favour of OS server operating system .... 35Figure 5–3 Usage of different OS databases........................................................................... 38Figure 5–4 Importance of all criteria for decision in favour of OS database ........................... 39Figure 5–5 Usage of different OS desktop software ............................................................... 41Figure 5–6 Importance of all criteria for decision in favour of OS desktop software .............. 42Figure 5–7 Usage of different OS software in connection with websites ................................ 44Figure 5–8 Importance of all criteria for decision in favour of OSS in connection with websites ....................................................................................................... 46Table 5–1 Establishment characteristics influencing choice of OSS as server operating system .................................................................................... 50Table 5–2 Establishment characteristics influencing choice of OSS for databases ........................................................................................................ 51Table 5–3 Establishment characteristics influencing choice of OSS on desktop computers ......................................................................................... 52Table 5–4 Establishment characteristics influencing choice of OSS in connection with websites................................................................................. 53


9

1 Introduction

1 Introduction

Report is part of FLOSS project

This report constitutes the first part of the final report from the FLOSS project,1

which was conducted from June 2001 – June 2002 by Berlecon Research and the In-ternational Institute of Infonomics at the University of Maastricht. The project wasfinanced by the European Commission under the Information Society Technologies(IST) thematic programme.

Project provides data on OSS usage

Purpose of the FLOSS project was – among other things – the collation of a base ofhard data on the importance and role of Open Source and free software in today'seconomies as well as an impact assessment for policy and decision-making. This re-port provides such data in the form of the results from a survey about the use of OpenSource software (OSS)2 in European enterprises and public institutions, about theirmotivations for using OSS, and about the benefits they derive from its use.

Fieldwork in spring 2002 with 395 detailed interviews

From February to May 2002 the fieldwork for the FLOSS user survey was conducted.Altogether 1,452 companies and public institutions in Germany, Sweden and UKwith at least 100 employees were asked by telephone whether they use Open Sourcesoftware. 395 of these were indeed using Open Source software in some way or an-other or were at least planning to do so within the next year. These establishmentswere interviewed in detail.

Two sorts of resultsThere are two sorts of results from this survey. First of all, the number of companiesin the different regions that use Open Source software provides some informationabout the actual popularity of OSS within these establishments. As this survey wasonly conducted at the establishment level, the focus is on deliberate economically mo-tivated usage decisions and not on the use of OSS for ideological or other personalreasons.

And secondly, the answers to several detailed questions posed to those establishmentsactually using OSS provides more insight into the motivations for and benefits fromusing Open Source.

Some definitionsAs the survey has been conducted among enterprises as well as public sector institu-tions, we will call the survey units “establishments” to capture the for-profit as well asthe non-profit entities. We will call both “professional users” to distinguish themfrom private users of Open Source software.

Outline of the reportThis report is organised as follows: Chapter 2 describes the survey’s methodology aswell as its effectiveness. It also contains important information about how the surveyresults can be interpreted. Chapter 3 presents results about the spread of Open Source

1. Free/Libre Open Source Software: Survey and Study. More information about the project canbe found on the Internet at the following sites: www.infonomics.nl/FLOSS / and www.ber-lecon.de/services/FLOSS/.

2. We have restricted this survey to Open Source software to obtain clear conclusions. Includingall other sort of free software as well as those software pricing schemes where some softwareparts are freely available would only lead to confusion. As the survey results show, the price tagdoes play an important role for the decision, though.


1 Introduction

10

usage within the three countries in general as well as for specific usage areas. Chapter4 contains survey results about the attitude of Open Source using establishments to-wards Open Source. And finally, chapter 5 contains survey results about the benefitsOSS users derive from this sort of software in four different usage areas.


11

2 Survey methodology


Two major objectives …In designing the survey of professional OSS users, two major objectives had to be met.First of all, the survey should provide fairly accurate information about the use ofOSS in general as well as in several popular application areas (e.g., as server operatingsystem). And secondly, the survey should provide reliable information about the mo-tivation for and benefits from using Open Source software.

… and a two-step survey approach

To obtain this information, a two-step approach was chosen, where establishmentswere first asked whether they are currently using OSS or planning to do so within thenext year, together with a short definition of Open Source software. Altogether 1,452establishments were asked this question. If they answered with yes, as 395 did, theywere given a detailed set of questions. Within these establishments, the target personwas the person responsible for IT decisions and administration. The person shouldbe able to answer questions about the establishment’s IT decisions and in additionhave a basic understanding of the technical issues.

Survey conducted by telephone

The professional user survey was conducted by telephone from February to May2002 by PbS AG from Munich. Telephone interviews have been chosen, as it is ratherdifficult to reach the targeted IT decision makers in other ways. People being respon-sible for IT related issues in commercial establishments currently belong to the mostoften interviewed professionals and are therefore typically reluctant to participate ifthey are not directly contacted.

Survey sample

Regions Germany, Sweden and the UK

The survey was intended to yield information about OSS use in several countries ofthe European Union. Due to budgetary restrictions, interviews could only be con-ducted for a limited number of countries. We have chosen Germany, Sweden and theUK. While the first and the last represent significant markets in the European Union,the second is a typical case for a small country, which has in addition a high IT usagerate. Furthermore, especially Germany and Sweden were of interest, as desk researchrevealed that they show opposite extremes of OSS usage: According to the last Inter-net Operating System Counter from April 19993 in Germany 42,7% of Internethosts were running Linux, while the same figure for Sweden was only 16,9%.

Survey stratification by two size classes …

To be able to compare the survey outcome by region, size or industry, the sample wasstratified by eight strata or quota. Country, establishment size and industry were cho-sen as characteristics for determining to which stratum an establishment belonged.Indicator for size was the number of employees per establishment. Entities with lessthan 100 employees were not included in the sample.

3. Leb.net/hzo/ioscount/.



12

❑ Quota one contains establishments with 100 to 500 employees per unit.❑ Quota two contains establishments with more than 500 employees per unit.

… and four industries In addition, there were four sample quotas based on industries. We distinguish be-tween the public sector and three quotas of the private sector. The private sector quo-tas were differentiated according to the amount of IT spending in relation to therevenues per industry. Motivation for this stratification was that industries with ahigh IT intensity – and thus high IT expenditures – might be more familiar with OSSand might therefore show a different usage pattern from those with lower IT spend-ing ratios.4

High IT intensity Quota one includes private industries with a high IT intensity. The IT spending inrelation to revenues is 4.1 percent or higher. According to the NACE classification,5

industries in this quota are:

❑ NACE J: Banking, insurance, and other finance❑ NACE I 64: Communications industry❑ NACE I 60-63: Transport industry❑ NACE K: Business services❑ NACE N: Health industry

Medium IT intensity Quota two includes private industries with medium IT intensity. The IT spending inrelation to revenues is a minimum of 2.8 percent but smaller than 4.1 percent. Ac-cording to the NACE classification, industries in this quota are:

❑ NACE E: Utilities❑ NACE DA-DJ: Process manufacturing, which includes all industries that

transform raw materials into products or into substances with new physicaland chemical properties (food, drink and tobacco, textile and leather, woodand fibre, paper and paper products, chemicals, rubber and plastics products,preliminary processing of non-metallic mineral products and metals)

❑ NACE DK-DN: Discrete manufacturing, which includes industries thattransform semi-finished products into final products (machinery and equip-ment not elsewhere classified, electrical and optical equipment, transportequipment, furniture, and recycling)

Low IT intensity Quota three includes private industries with low IT intensity. The IT spending in re-lation to revenues is lower than 2.8 percent. According to NACE, industries in thisquota are:

❑ NACE G 51: Wholesale trade❑ NACE G 50 and G 52: Retail and automobile trade❑ NACE H: Hotels and tourism❑ NACE F: Construction Industry

Public sector Quota four includes the public sector. According to the NACE classification, we de-fined it as:

❑ NACE L: Government❑ NACE M: Education

Because of supposedly low use of Open Source software agriculture, mining and otherservices were not included in the survey. According to NACE classification these areNACE A-B (Agriculture), NACE C (Mining) and NACE O (Other Services).

4. The usage ratios were taken from META Group and Rubin (2001): IT Spending as a % of Rev-enue, www.metricnet.com.

5. NACE = Nomenclature statistique des Activités économiques, official classification of indus-tries used in the European Union.


13


Table 2–1Survey observations per strata

Observation matrix for survey

This stratification of the sample can be summarised in a matrix. Each cell in the ma-trix contains the number of observations in the sample for establishments with thesecharacteristics. These cells do not show identical numbers of observations for a varietyof reasons. First of all, due to budgetary constraints and high refusal rates, fewer in-terviews were conducted in the UK than in the two other countries. Secondly, insome cells the number of establishments in the universe was rather small, the refusalrates were higher than on average and/or the usage of Open Source software was lesswidespread. In one extreme case a combination of these led to the situation that onlya single observation exists.

Participation and refusal

IT decision makers were difficult to reach

Participation in this survey was rather uneven, with relatively high refusal rates inmany categories. As stated above, at the time of conducting this survey IT decisionmakers were difficult to reach. Not only had they been surveyed many times duringthe previous e-commerce boom, but they were in addition busy coping with tight-ened budgets and with restructuring their establishments e-business activities.

Refusal rates differ between countries and industries

As table 2–2 shows, the refusal rates differ between countries as well as between in-dustries and size classes, ranging from as low as 5.6% (large public sector establish-ments in Sweden) to as high as 91.5% (large medium IT intensive companies inGermany).

Assumption of equal OSS usage patters between participants and refusals

A major assumption for the subsequent analysis is that the OSS usage patterns ofthose that refused outright, do not differ from those that took part. If this is not thecase, the OS usage would most likely be higher with those companies that participat-ed than with those that refused, biasing OSS usage rates upwards. Justification behindthis guess is that those not using OSS might more likely refuse when they hear thetopic of the survey as they consider it to be not of interest.

Table 2–2Refusal rates in survey

UK Sweden Germany Total

small large small large small large

High intensity (NACE I,J,K,N) 7 20 20 20 20 20 107

Medium Intensity (NACE D, E) 9 1 20 20 20 20 90

Low Intensity (NACE F, G, H) 7 7 21 14 21 19 89

Public sector (NACE L, M) 20 13 20 16 20 20 109

Source: Survey results.

UK Sweden Germany


High intensity (NACE I,J,K,N) 86.5% 57.8% 62.7% 50.2% 76.6% 83.3%

Medium Intensity (NACE D, E) 72.3% 83.8% 42.9% 22.8% 69.2% 91.5%

Low Intensity (NACE F, G, H) 80.3% 87.0% 21.4% 31.7% 79.7% 88.1%

Public sector (NACE L, M) 53.4% 72.4% 29.9% 5.6% 77.9% 63.3%

Total 76.2% 39.8% 82.0%

Source: Survey results.



14

Weighting scheme

To be able to draw conclusions about the use of Open Source software in the differentcountries, weights have to be calculated for each cell to bring the sample distributionin accordance with the real distribution of establishments across these cells. Other-wise the stratification of the sample would lead to distortions. E.g., cells with a highnumber of observations in the sample but a low number of establishments in realitywould have a larger weight in our sample than in reality. If they differ significantlyfrom the remaining ones, the results could be significantly distorted without weight-ing.

Weights derived fromaddress broker data

To calculate these weights, we used distributions of addresses over size classes and in-dustries from major address brokers in the surveyed countries, as no detailed officialstatistics exist. The brokers chosen were Bertelsmann/Creditreform in Germany,Dun & Bradstreet in the UK and PAR-Guiden for Sweden. These weights are useddirectly in the calculation of the Open Source penetration in chapter 3. The resultsin that chapter are thus representing the universe of all establishments with 100 ormore employees within the industries investigated.

Further weighting by OSSpenetration rates

For all other calculations, the results are further weighted by the OSS penetrationrates in the different cells. Thus, all information in chapters 4 and 5 is representingthe universe of OSS using establishments with 100 or more employees within the in-dustries investigated in each of the three countries.

Countries normalised toidentical size

To avoid combined results to be dominated by Germany – which has about ten timesthe size of Sweden – we normalised the weights to identical country sizes. Thus theSwedish establishments using OSS have the same weight in the combined sample asUK or German establishments. The pooled results can then be interpreted as averagesof the countries.

Questionnaire

Extensive pre-testingperiod

The survey questionnaire was prepared as a result from previous work done in theFLOSS project. In preparing the questionnaire, we had numerous interviews withrepresentatives from establishments that use Open Source software with the purposeto find out which questions can and which cannot be asked in such a survey. Severalpre-tests were conducted with preliminary versions of this questionnaire. One of theresults of these tests was that most representatives could not give any detailed mone-tary figures for the benefits they derived from using Open Source software. Thereforethe questions were posed in a different way. More detail is given in chapter 5.

Questionnaire reprinted inappendix

The full questionnaire in the English version is reprinted in the appendix to this re-port. Establishments were surveyed in their local language.


15

3 Professional use of Open Source software

3 Professional use of Open Source software

Measuring OSS professional use poses special challenges

Measuring the professional use of Open Source software in companies and public in-stitutions is more difficult than measuring the use of commercial software. As OSScan be freely duplicated, the number of sales by distributors, by hardware manufac-turers that have pre-installed this software, as well as by Internet sites offering thissoftware for download does not correlate in a reliable way with the actual number ofinstallations.

On the one hand, as OSS can be freely copied, one would assume that the actualnumber of installations is higher than the number of copies sold or downloaded. Buton the other hand, as OSS is typically (almost) free, there might be many purchases,distributions on magazine CD-ROMs or downloads that lead to a short installationfor testing purposes, followed by deletion. One would intuitively assume the first ef-fect to be larger, but there is – at least to our knowledge – no reliable estimation ofthe latter’s size. The whole issue is further complicated by the fact that neither CD-ROM distributions nor downloads (and also not some surveys)6 distinguish betweenprivate and professional users.

Usage number by market research companies differ considerably

As a consequence of these problems, even the usage numbers provided by professionalmarket research companies like Forrester, Gartner, IDC or META Group differ. Inaddition, much of the research has further problems: Most work has been publishedin 1999 or 2000 and might not be accurate any more. Also one cannot exclude thatthe Internet hype of 1999 and 2000 has biased some of the results, especially thoselooking into the future. And finally, many of the studies are focused on Linux and donot provide much information about the professional use of other forms of OSS.7

How large the differences between estimations are can be shown using Linux as anexample: In summer 2001 Gartner stated that Linux was installed on almost 9% ofthose servers shipped in the third quarter of 2000. At the same time, however, IDCstated that Linux already constituted around a third of the server market.8 Accordingto IDC, in summer 2001 already 40% of companies in the US and Europe were al-ready using Linux on test or production systems.

Caveats for careful investigations

These differences show that one has to take into account whether the respective num-bers describe software installations shipped with the hardware (especially importantfor operating systems like Linux) or whether they also include software installed later.The latter can only be obtained by asking the respective users directly. Also, one has

6. Example is a survey conducted on the website of the German IT magazine c’t in June 2001.9,213 respondents filled out a questionnaire about their use of Linux. How important profes-sional users were is unclear. Cf. Diedrich, Oliver (2001): Und was machen Sie mit Linux?, c’t17/01, 186-189.

7. An overview can be found in Schmitz, Patrice-Emmanuel (2001): Study into the use of OpenSource Software in the Public Sector, Part 2: Use of Open Source in Europe.

8. Computerwoche (2001): IDC veröffentlicht optimistische Linux-Prognose, 16.8.2001.


3.1 Professional use of Open Source software

16

to disinguish between percentages of compaies and percentages of server installations.Finally, one has to distinguish between the operational use of Open Source softwareand the installation for testing purposes. These potential problems have been takencare of in the FLOSS professional user survey.

The following section 3.1 contains information about the use of Open Source soft-ware in establishments in Germany, Sweden and UK in general. Section 3.2 goes intomore details and presents results about the usage of Open Source software in differentsoftware application areas.

3.1 OSS use in general

Considerable usagedifferences among

countries

Figure 3–1 shows that the usage of Open Source software differs considerably amongthe three countries investigated. While 43.7% of establishments in Germany are us-ing OSS, only 31.5% of British establishments and only 17.7% of Swedish establish-ments do so.

Figure 3–1Usage of Open Source

software by country

For Germany and Sweden these numbers fairly accurately replicate those obtained bythe Internet Operating System Counter (IOSC) in 1999.9 These project calculatedLinux to be running on 42.7% of hosts in Germany and on 16.9% of hosts in Swe-den. For the UK, the IOSC figured Linux to be running on 24.3% of hosts. Thus,the correlation is not as strong as for Sweden and Germany, although the ranking isthe same.

Also differences withrespect to size

As table 3–1 shows in more detail, usage rates not only differ by country, but alsowithin countries. For example, the OSS usage rates of larger establishments are higherthan those of small establishments in 8 of the 12 cells. This result is plausible sincelarge establishments typically have a more diverse IT infrastructure increasing theprobability that for some purpose OSS is being used. One would therefore expecthigher OSS usage rates in these establishments.

9. Leb.net/hzo/ioscount/

UK

Sweden

Germany

0% 20% 40%B

erle

con

Rese

arch

200

2

n=1,452

43.7%

17.7%

31.5%


17

3.2 OSS use by IT area

Table 3–1Current and planned professional use of OSS in Germany, Sweden and UK

Differences with respect to country and sector

Quite consistently observable are above-average OSS usage rates in the public sector.In 5 out of 6 cells is the OSS usage rate higher in the public sector than on average inthe respective country. There are also differences between the usage rates in the threedifferent private sector segments. These are, however, not in any way systematicacross countries and size classes.

Results have to interpreted with appropriate care

The highest usage rate across all cells could be observed in large companies with highIT intensity in the UK. 74.1% of those companies contacted stated that they are us-ing Open Source software. The lowest rate – not counting the cell with only one OSSusing observation – could be observed within large companies with high IT intensityin Sweden (13.2%). This shows that one has to be very careful with generalisationsabout what sort of establishment tends to use OSS.10

One also has to be careful in interpreting the results as they rest on the assumptionthat those establishments that refused to participate and those that could not bereached differ not significantly in their OSS use from those surveyed. Also it has tobe considered that for the large UK companies with medium technology use only asingle OSS using observation exists.

3.2 OSS use by IT area

OSS as server OS most popular (15.7%)

Table 3–2 shows the average percentage of establishments using Open Source soft-ware in the four different IT areas they were asked about. Most popular is the use ofOSS as server operating system: On average 15.7% of establishments either currentlyuse Open Source software like Linux or Free/Open BSD for server operating systemsin regular IT operations or are planning to do so within the next year. As one can see,the differences between countries are considerable. While 30.7% of German estab-lishments employ OSS this way, only 10.1% of Swedish and 6.4% of British estab-lishments do.

UK Sweden Germany


High intensity (NACE I,J,K,N) 25.0% 74.1% 20.4% 13.2% 27.0% 51.3%

Medium Intensity (NACE D, E) 39.1% 9.1% 14.6% 32.8% 45.5% 51.3%

Low Intensity (NACE F, G, H) 25.0% 14.3% 13.6% 20.3% 52.8% 44.4%

Public sector (NACE L, M) 32.8% 38.2% 16.4% 23.5% 44.4% 69.0%

Total 31.5% 17.7% 43.7%

Source: Survey results (n=1,452).

10. The same applies to cross-country generalisations. For example, it is sometimes argued that the(license-fee-free) Open Source software can help poorer countries in setting up their IT infra-structure. However the IOSC shows commercial software to have been the most popular hostoperating system in many poor countries in 1999. Only in some countries was Linux on posi-tion one.


3.2 Professional use of Open Source software

18

Table 3–2Current and planned

professional use of OSS indifferent application areas

Next are OSS databases(11.1%) …

Next in popularity is the use of OSS for databases. MySQL, PostgreSQL, Interbaseor SAP-DB are examples of such Open Source software. On average 11.1% of the es-tablishments employ OSS for databases. In this area, the differences are less pro-nounced. The OSS usage rate in Germany (15.7%) is slightly more than twice as highas in Sweden, where it is lowest with 7.6%.

… and then OSS for creatingand operating websites

(10.1%)

On average 10.1% of the establishments use OSS in connection with creating or op-erating websites. There is a large variety of applications that are used in this area, e.g.Apache, PHP, Perl, Python, Squid or Open Source content management systems.Again the usage rate is highest in Germany (16,2%) and lowest in the UK (6.5%).11

OSS on desktops is not verywidespread (6.9%)

Finally, Open Source software can also be used on desktop computers. Examples areLinux as a desktop computer operating system, desktop extensions like KDE orGnome but also application programs like Mozilla or StarOffice/OpenOffice. How-ever, OSS is not used very frequently on desktops. On average only 6.9% of the es-tablishments in the three countries investigated use OSS on desktops – and this doesnot mean that they use OSS on all their desktops. Again, the usage is highest in Ger-many (12%) and very low in Sweden, where only 3.3% of establishments use OSSon some of their desktop computers.

OSS use on desktops morefrequent in smaller

establishments

In all usage areas except the use of OSS on desktops there is no clear indication acrosscountries of a higher Open Source software usage rate by smaller or by larger estab-lishments. Only the use of OSS on desktops is more frequent in smaller than in largerestablishments (in Sweden not significantly).

UK Sweden Germany Total


OSS as server operating system 8.1% 3.7% 9.8% 11.0% 30.7% 30.6% 15.7%

6.4% 10.1% 30.7%

OSS for databases 13.3% 4.6% 7.5% 8.2% 14.1% 20.8% 11.1%

9.9% 7.6% 15.7%

OSS on the desktop 7.6% 2.0% 3.4% 3.2% 13.7% 6.5% 6.9%

5.4% 3.3% 12.0%

OSS for websites 7.9% 4.3% 7.5% 8.7% 15.8% 17.3% 10.1%

6.5% 7.8% 16.2%

Source: Survey results (n=395).

11. These numbers seem to be rather low when taking into account the huge popularity of Apacheas a web server. This might be due to many establishments not hosting their websites them-selves.


19

4 Attitude to and importance of Open Source software in general

4 Attitude to and importance of Open Source software in general

4.1 Attitude to Open Source software

General attitudes to OSS vs. specific buying decisions

An enterprise’s decision to use OSS can be driven by two sorts of motives. The firstsort of motives is application specific, e.g. an expected greater stability or lower costsfor that specific application in comparison to its commercial alternatives (cf. chapter5). The second sort of motives is more general, like the wish to support the OpenSource community by using Open Source software or by letting one’s IT personnelwork on OS projects on company time.

Seven statements about establishments’ attitude to OSS

To find out how important these motives are, we gave the surveyed establishmentsseven statements concerning general Open Source policies and usage motives andasked them, to what extend they agreed to these statements. One of these questionswas concerned with the possibility for software developers to work on Open Sourceprojects within their working time. The answers to these questions not only provideinformation about the establishments’ position to Open Source software but alsoabout the amount of (indirect) support that OS projects obtain from companies andpublic institutions.

The following statements were provided:

❑ We use Open Source Software because we want to be more independent fromthe pricing and licensing policies of the big software companies.

❑ By using Open Source Software we want to support the Open Source com-munity.

❑ We use Open Source Software because IT specialists for this kind of softwareare more easily available on the labour market than specialists for proprietarysoftware.

❑ We prefer using Open Source Software – that’s part of our company policy.❑ Our software developers are free to work on Open Source projects within their

time at work.❑ We are deliberately working together with Open Source service companies in

order to support the development of Open Source software.Regression analysis for each question

To find out whether there are specific characteristics of an establishment that influ-ence the attitude towards Open Source, regression analyses have been conducted. Wehave tested for the explanatory power of the country, the size class, the industry, theshare of IT personnel in total employees as well as the number of PCs per employee.While the detailed statistical results are provided upon request, this section discussesthose effects that have been found to be statistically significant.


4.1 Attitude to and importance of Open Source software in general20

Independence from pricing and licensing policies of big software companies

Microsoft’s licensingchanges might intensify

wish for independence

One argument in favour of using OSS is to become (more) independent from thepricing and licensing policies of big software companies. This is a major point em-phasised by the proponents of OSS in public administration as, for example, the re-cent discussion in Germany has shown. With budgets tight, previous softwareversions (e.g. Windows NT) becoming unsupported and new licensing schemes com-ing up, which are considered to lead to higher software expenditures, establishmentsmight wish to become less dependent from pricing and licensing policies of big soft-ware companies.

Figure 4–1Independence from bigsoftware companies as

motivation

Almost 56% agree tostatement

As figure 4–1 shows, independence from the pricing and licensing policies of big soft-ware companies is a major motivation for those establishments that already use OSS.On average over the three countries, almost 56% of those establishments that use OSsoftware either agree totally or somewhat to this statement. On average, the respond-ents answered between “somewhat agree” and “neither nor”.12 However, there is alsoa strong group of almost 29% who disagreed with this statement. Thus, most estab-lishments have a clear position on this issue.

Outcome of regressionanalysis

The regression analysis shows the following statistically significant relationships be-tween basic characteristics of an establishment and the level of agreement to the state-ment:

❑ Small establishments agree significantly less to this statement than large estab-lishments.

❑ German establishments agree significantly less to this statement than UK es-tablishments.

❑ There is no evidence for differences in answering this question between pri-vate and public sector nor is there among the different private sectors.

totally agree (30%)

somewhat agree (26%)

neither nor (12%)

somewhat disagree (10%)

totally disagree (19%)

no answer/ don't know (4%)

Ber

leco

n Re

sear

ch 2

002

n=395

"We use Open Source Software because we want to be more independent from thepricing and licensing policies of big software companies."

12. The answers are coded from 1 = totally agree to 5 = totally disagree. The mean for this questionis 2.8.



Figure 4–2Differences in importance of independence from big software companies

Strong disagreement in German establishments

As figure 4–2 shows, support was strongest in the UK, where 46% of OS softwareusers totally agreed to this statement and 62% agreed at least to some extent. In con-trast, only 55% of the German respondents agreed either totally or somewhat to thestatement. What is more important, is that Germany also shows the strongest disa-greement with more than 22% of respondents who totally disagree. For the UK, thisvalue is only slightly above 13%.

For larger establishments, independence was more often important for OSS use

Considering the establishment size, the larger entities in the survey show a highershare of respondents that totally agree to this statement as well as a lower share ofthose that totally disagree. A reason for this outcome might be that larger establish-ments typically spend more on software. Therefore the savings from lower prices ormore favourable licenses can be substantial in these companies.

Support of the OSS community

About one third wants to support OS community by using OSS

Some establishments might use Open Source software to support the Open Sourcecommunity. While one would consider such an effect to be more important with pri-vate OSS users than with professional users one cannot exclude a priori that it playsa role. As figure 4–3 shows, this attitude towards OSS is not very wide spread but ex-ists. On average almost 14% totally agree that by using OSS they want to support theOpen Source community, and almost 21% agree at least somewhat. Taking into ac-count that this survey is about professional users, these values are astonishingly high.They are, however, lower than for the previous statement, as the average answer isslightly more negative than “neither nor”.

*

Large

Small

Germany

UK

Total

0% 20% 40% 60% 80%

35% 27%

28% 25%

25% 30%

46% 16%

30% 26%

totally agree somewhat agree

Ber

leco

n Re

sear

ch 2

002

"We use Open Source Software because we want tobe more independent from the pricing and licensingpolicies of big software companies."



Figure 4–3Wish to support the OS

community as motivation


The regression analysis shows the following relationships between basic characteris-tics of an establishment and the level of agreement to the statement:

❑ While UK and Sweden do not differ significantly, German establishmentsagree significantly less to this statement than UK as well as Swedish establish-ments do.

❑ There are no significant differences between size classes.❑ There is no evidence for differences in answering this question between pri-

vate and public sector nor among the different private sectors.Figure 4–4

Differences in importanceof wish to support the OS

community

totally agree (14%)


neither nor (18%)somewhat disagree (12%)



Ber

leco

n Re

sear

ch 2

002

"By using Open Source Software we want to support the Open Source community."

n=395

*

Germany

Sweden

UK

Total

0% 20% 40% 60% 80%

9% 20%

14% 22%

26% 18%

14% 21%

totally agree somewhat agree

Ber

leco

n Re

sear

ch 2

002

"By using Open Source Software we want to supportthe Open Source community."



UK establishments have strong wish to support OSS community

The wish to support the Open Source community is a rather strong driver for the useof OS software in the UK. 44% of the surveyed establishments agreed to this secondstatement at least somewhat. In Sweden (36%) and Germany (29%) the degree ofagreement was considerably lower. Also disagreement with the statement was ratherstrong in Germany.

More easily available IT specialists

Do widespread use of Linux and OSS at universities lead to many OSS specialists?

Especially during the heyday of the New Economy, availability of computer special-ists was a serious problem in many companies and institutions. As some Open Sourcesoftware is by now relatively widespread (especially Linux) and as students often getfamiliar with Open Source software during their university education, specialists forOS software might be more easily available than specialists for commercial software– so far the hypothesis. To test it, the respondents where asked whether they use OSSbecause IT specialists for this kind of software are more easily available on the labourmarket than specialists for proprietary software.

No evidence for better availability of IT experts being a strong motivation for OSS use

The answers to this statement show that the hypothesis has to be rejected. The meanresponse was less negative than “somewhat disagree”. On average only 21% of the re-spondents agreed at least somewhat to this statement. At the same time 45% totallydisagreed, meaning that either the better availability of OSS specialists does not playa role for their decision towards using OSS or that they do not think that there existssuch an advantage. As only OSS users were surveyed in detail, we cannot answer thequestion whether maybe even a perceived shortage of OSS specialists exists that keepsestablishments from employing Open Source software. The results suggest, though,that this question might well be worth investigating as it also bears consequences forpolicy that wants to support the use of Open Source software.

Figure 4–5Availability of IT specialists as motivation

totally agree (7%)


neither nor (15%)




Ber

leco

n Re

sear

ch 2

002

n=395

"We use Open Source Software because IT specialists for this kind of software aremore easily available on the labour market than specialists for proprietary software."



Figure 4–6Differences in importance

of availability of ITspecialists



❑ There is weak evidence for a higher share of IT workers in employees leadingto less agreement with the statement. This might point to available knowledgein-house and therefore no greater need to consider the availability of specificknowledge on the labour market.

❑ There are clear country differences. German establishments tend to disagreemore with the statement than Swedish as well as UK establishments. There isno significant difference between the latter two, though.

❑ There are no significant differences between small and large establishments aswell as between the different sectors.

Strongest disagreementwith statement to be found

in Germany

As figure 4–6 shows, in Germany 71% of respondents disagreed at least somewhat tothis statement, most of them even totally. In Sweden, where a large part neitheragreed nor disagreed, disagreement was still 41%. At the same time about 29% ofSwedish establishments agreed to this statement at least to some extent. Again thismight be due to differences in the availability of OSS specialists on the different coun-tries’ labour markets or to differences in the importance such availability has for theestablishments in the different countries.

-80% -60% -40% -20% 0% 20% 40% 60% 80%

-60% -11% 10% 3%

-32% -9% 19% 10%

-42% -12% 15% 10%

-45% -10% 14% 7%

somewhat disagree totally disagree somewhat agree totally agree

Ber

leco

n Re

sear

ch 2

002

Total

Sweden

UK

Germany

"We use Open Source Software because IT specialists for this kind ofsoftware are more easily available on the labour market thanspecialists for proprietary software."

totally disagree somewhat disagree



OSS as company policy

Company policies keep IT infrastructure homogeneous and help to achieve company goals

As the German discussion about the use of OSS in public administration shows, com-panies and institutions might develop internal policies about which software to use.Such policies help in keeping the IT infrastructure homogeneous and thereby reducecosts. But they can also be motivated by other, strategic reasons. For example, an es-tablishment might have the policy to use mostly OSS to save software license fees. Orit might have the policy to use OSS due to a perceived better overall security or be-cause it wants to be able to modify the source code eventually. Thus, there are severalpossible motivations for such a company policy in favour of Open Source software.

Figure 4–7OSS use as company policy

Open Source company policy not of large relevance

A dedicated company policy to prefer using Open Source software is only in very fewestablishments a motivation for using OSS. On average only 19% of establishmentsagreed to this statement at least somewhat while 48% totally disagreed. The mean isslightly less negative than “somewhat disagree”. It is, however, even less affirmativethan the answer to using Open Source software as company policy.

Outcome of regression analysis


❑ Companies from sectors with typically low IT intensity tend to agree more tothe statement than public institutions. Their assessment does, however, notdiffer significantly from that in other private sectors, nor do the other privatesectors differ from the public sector.

❑ Again Germany is different. German establishments tend to disagree morestrongly with this statement than Swedish or UK establishments.

❑ There is no significant influence of a company’s size or its IT infrastructure.

totally agree (8%)


neither nor (15%)




Ber

leco

n Re

sear

ch 2

002

"We prefer using Open Source Software – that’s part of our company policy."

n=395



Figure 4–8Difference in relevance of

OSS use as company policy

OSS as company policymost strongly in the UK –

strong disagreement inGermany

Support to this statement is strongest in the UK – 24% agree at least to some extent– followed by Sweden (23%). In Germany only 12% agreed at least to some extentto the statement that they use OSS because it is company policy. At the same timethe fraction of those totally disagreeing is highest in Germany (51%). A further 16%of German establishments disagree at least to some extent. As figure 4–8 also shows,the significant difference of answers by companies from sectors with low IT use is dueto their low fraction of disagreement with this statement – 40% vs. 48% on averagetotally disagree – not due to their strong agreement.

OSS development during work time allowed

FLOSS developer surveypoints at OSS development

during work time to beimportant

It is often said that much of Open Source software is being developed during worktime, as developers are allowed to (or simply do) work on OSS projects during theirwork time. To some extent this is supported by the FLOSS developer survey, whichindicated that 29% of OS programmers are paid for developing Open Source or freesoftware while 24% are not paid but can develop OS/FS at work.13 One would as-sume especially the latter to be the case more often at public institutions (e.g. univer-sities). While we do not have an answer for the relevance of OSS development duringwork time that is not allowed, we asked the responsible persons whether their soft-ware developers are free to work on Open Source projects within their time at work.

Bimodal distribution with36% allowing OS

development at work

Figure 4–9 shows this freedom to be indeed of importance. On average 36% of re-spondents agreed at least somewhat to this statement, 19% of them totally. As expect-ed, the distribution is quite bimodal – either working on OSS is allowed or not. So45% of respondents disagree at least somewhat, most of them totally. The mean isslightly more negative than “neither nor”.

-80% -60% -40% -20% 0% 20% 40% 60% 80%

-40% -16% 10% 9%

-51% -16% 6% 6%

-47% -9% 14% 9%

-44% -15% 15% 9%

-48% -13% 11% 8%


Ber

leco

n Re

sear

ch 2

002


Total

Sweden

UK

Germany

Low IT Intensity

"We prefer using Open Source Software – that’s part of our companypolicy."

13. Ghosh, Rishab Aiyer; Ruediger Glott; Bernhard Krieger; Gregorio Robles (2002): FLOSSDeveloper Level Analysis, working paper, mimeo.



Figure 4–9OSS development during work time allowed

Outcome of regression analysis


❑ Companies from the sector with on average high IT intensity tend to agreemore with the statement than companies with on average medium IT inten-sity. This probably reflects software development being part of the high IT in-tensity sector. There is also weak evidence for a similar relationship of high ITintensity companies and low IT intensity companies.

❑ German establishments tend to disagree more with this statement than Swed-ish companies.

Figure 4–10Differences in relevance of

OSS development duringwork time

totally agree (19%)


neither nor (8%)




Ber

leco

n Re

sear

ch 2

002

"Our software developers are free to work on Open Source projects within their time atwork."

n=395

-80% -60% -40% -20% 0% 20% 40% 60% 80%

-21% -9% 26% 21%

-51% -9% 13% 18%

-47% -8% 11% 16%

-49% -10% 17% 9%

-25% -9% 17% 29%

-37% -9% 17% 19%


Ber

leco

n Re

sear

ch 2

002


Total

Sweden

Germany

Low IT Intensity

Medium IT Intensity

High IT Intensity

"Our software developers are free to work on Open Source projectswithin their time at work."



Swedish OS-usingcompanies are more liberal

than German ones

While on average 36% of respondents agreed to this statement, the correspondingpercentage in Germany was only 26%. At the same time many more establishments– 59% vs. 46% on average – disagreed at least somewhat to the statement. In com-parison to the Germans, Swedish establishments that use OSS are much more liberalwith respect to letting employees work on OS projects on company time.

Large fraction of high ITintensity companies

supports OS developmentat work

Figure 4–10 shows also a rather strong support of OSS development by companiesfrom sectors with on average high IT intensity. 47% agree at least somewhat to thestatement. In the low IT-intensity sector, this share is lowest with 27%. At the sametime, total disagreement is only at 21% in the high IT intensity sector – much lessthan the average of 37% – but rather high (51%) in the medium IT intensity sector.

Using OS service companies to support OS development

During the late 1990s, many Open-Source-oriented start-ups were founded that triedto further the development of Open Source software and at the same time to makemoney from selling services. These ranged from installation support to developmentof complex enterprise solutions on the basis of Open Source software. Since then,many of these companies have failed. Nevertheless, it seemed to be of interest to askthose companies using Open Source software about whether they are deliberatelyworking together with Open Source service companies in order to support the devel-opment of Open Source software.

Strong disagreement As figure 4–11 shows, this seems to be the case only to a very limited extent. On av-erage only 15% of the OS using companies in the three countries agree to this state-ment at least somewhat. With a mean of statements around “somewhat disagree”, thisstatement received the strongest disagreement of all.

No significant differencesdue to establishment

characteristics

The regression analysis shows no significant differences in answering this question.Neither size nor industry, country, share of IT users, or PC intensity imply a differentassessment of this issue. OS service companies simply cannot count on any particularbonus by anyone for supporting OS development in some way or another. That issomething the dotcoms in this field had to experience, too, during the last two years.

Figure 4–11Using OS service

companies to support OSdevelopment

totally agree (6%)

somewhat agree (8%)

neither nor (11%)

somewhat disagree (15%)totally disagree (52%)


Ber

leco

n Re

sear

ch 2

002

n=395

" We are deliberately working together with Open Source service companies in order tosupport the development of Open Source software."



Summary

Establishments agree most strongly to wish to become independent from large software companies

Figure 4–12 shows the weighted mean of answers to the different statements. As thefigure shows, agreement is strongest with the statement that establishments use OSSto become more independent from the pricing and licensing policies of large softwarecompanies. The average answer is between “agree somewhat” and “neither nor”.

Figure 4–12Attitude to Open Source software in general

Altruistic motives are of lesser importance …

Next in order are different ways to support the OS community, either indirectly byusing OSS or directly by letting one’s developers work on OSS development on com-pany time. Nevertheless, these assessments are already more on the negative side,which shows that individual gains for the establishments are a much more importantreason for using OSS than the altruistic wish to further OSS development or to sup-port the OS community.

… as are labour market considerations, company policy or cooperation with OSS service companies

The least agreement was on average found with the statement that companies mightuse OSS because IT specialists are more easily available, with the statement that OSSuse might be company policy as well as with the statement that establishments mightwork together with OSS service companies to support OSS development.

Figure 4–13Attitude to Open Source software in German establishments in general

Strong disagreement of German companies due to different composition of German OSS users

German establishments tend to disagree more strongly than the average to all state-ments, as a comparison of figures 4–12 and 4–13 shows. This does not mean, how-ever, that German companies per se are less supportive of Open Source software. Onehas to keep in mind that OSS use is far more common in Germany than in the other

0 1 2 3 4 5

Ber

leco

n Re

sear

ch 2

002

Independence from pricing and licensingpolicies of big software companies

Support of the OS community

More easily available ITspecialists


OSS development during worktime allowed

Using OS service companies tosupport OS development

totallydisagree

neither nor

totallyagree

somehowagree

somewhat disagree

3.88

3.31

3.23

2.60

3.80

4.05 n=395

0 1 2 3 4 5

Ber

leco

n Re

sear

ch 2

002

Independence from pricing and licensingpolicies of big software companies

Support of the OS community

More easily available ITspecialists


OSS development during worktime allowed

Using OS service companies tosupport OS development

4.26

3.49

3.82

4.11

4.07

2.71

n=395

totallydisagree

neither nor

totallyagree

somehowagree

somewhat disagree



two countries investigated. Thus this observed difference in attitude towards OSSmight simply be due to the fact that a larger share of OSS users in Sweden and theUK are still highly emotional early OSS users, whereas many German OSS using es-tablishments are doing so for more pragmatic reasons.

4.2 Importance of OSS within the IT infrastructure

Importance of OSS within ITinfrastructure points to

value of OSS forestablishment

Somewhat related to the previous section on the general attitude towards OS softwareis the question discussed in this section. Establishments might use OSS in differentintensities. For some companies or public institutions, the usage of OSS is an impor-tant element in their IT strategy, and they have made a deliberate decision to useOSS. Others experiment with OSS on certain occasions or use this sort of softwareonly on some machines or for unimportant tasks. To find out which value OSS hasin the establishments surveyed, we asked them how important from their point ofview Open Source Software was for their establishment’s IT infrastructure. Again theanswers to this question refer to those establishments that either use OSS already orplan to do so within the next year.

For most establishments using OSS, it is an important part of their IT infrastructure.As figure 4–14 shows, for 10% of the OSS using establishments it constitutes a veryimportant part of their IT infrastructure and for additional 23% it is of high impor-tance. 44% consider it to be of low or very low importance. On average, the respond-ents consider OSS to be slightly less than medium important for their ITinfrastructure.

Regression with additionalvariables

To obtain information about the relevance of firm characteristics for this question,we have conducted again a regression analysis with the same variables as for the atti-tude towards OSS. In addition, we added dummies for the current or planned regularuse of OSS as server operating system, for databases, on the desktop and in connec-tion with creating or operating websites to the estimation.

Figure 4–14Importance of OSS within

IT infrastructure

very high (10%)

high (23%)

medium (21%)

low (26%)

very low (18%)

don´t know (1%)

Ber

leco

n Re

sear

ch 2

002

n=395

From your point of view, what is the importance of Open SourceSoftware for your company’s IT infrastructure? Is the importance ...


314.2 Importance of OSS within the IT infrastructure

Figure 4–15Differences in importance of OSS within IT infrastructure

The regression produced the following statistically reliable results:

❑ There exist again country differences. For German establishments OSS is aless important part of their IT infrastructure then for UK or Swedish estab-lishments. The latter, however, do not differ in a statistically significant way.

❑ There is no significant difference between the assessment of OSS importancebetween smaller and larger companies.

❑ Companies from the high IT intensity sector consider their OSS to be moreimportant than establishments from the public sector do. There are, however,no further distinctions across sectors.

No difference according to area of OSS use

❑ Establishments that currently use OSS in one of the four areas in regular IToperations or plan to do so within the next year consider their OSS to be amore important part of their IT infrastructure than those that use it only oc-casionally or in important cases. While this was to be expected given the ques-tion, it comes as a slight surprise that there is no statistically significantdifference between the IT areas. Thus the answer to the question does not dif-fer between those establishments, e.g., that use OSS on client computers andthose that are running OSS server operating systems.

German establishments consider OSS to be less important than UK or Swedish ones

As figure 4–15 shows, OSS is especially in Sweden an important component of therespondents‘ IT infrastructure. 43% consider it to be of high or very high importance.In comparison, the corresponding value for Germany (25%) is significantly lower.The importance of OSS for the IT infrastructure of German establishments is alsosignificantly lower than in the UK, where 30% consider it to be at least of high im-portance. Compared to both other countries the percentage of German establish-ments attaching a low or even very low importance to this aspect is high with 57%.Again the caveat applies that due to the comparatively high OSS penetration in Ger-many the composition of OSS users is simply different from that in Sweden and theUK.

-60% -40% -20% 0% 20% 40% 60%

-13% -26% 28% 14%

-29% -17% 20% 9%

-13% -27% 22% 8%

-22% -35% 17% 8%

-17% -15% 30% 13%

-18% -26% 23% 10%

low very low high very high

Ber

leco

nRe

sear

ch20

02

Total

Sweden

Germany

UK

Public Sector

High IT Intensity

lowvery low

From your point of view, what is the importance of Open SourceSoftware for your company’s IT infrastructure? Is the importance ...



For companies from thehigh IT intensity sector,

OSS is more important thanfor public institutions

Evaluating the outcomes by sector shows a significant difference between the high ITintensity and the public sector. While in the high IT intensity sector 42% considertheir OSS to be important, in the public sector only 29% do so. At the same time29% of the public sector establishments consider OSS to be of very low importancefor their IT infrastructure whilst only 13% of the companies from the high IT inten-sity sector give a similar assessment.


335 Benefits from using Open Source software

5 Benefits from using Open Source software

Hard data on the benefits from using OSS

A major aim of this survey was to generate hard data on the benefits companies andpublic institutions derive from using OS software. This chapter goes into detail andpresents those survey results that contribute towards this target.

Companies unable to state direct or indirect monetary benefits

While it would have been wishful to obtain information on the direct monetary valueof using OSS, it turned out in the pre-test for this survey that this was not a realistictarget. Those companies that were interviewed during that phase were generally un-able to provide even rough estimates about the monetary value derived from usingOpen Source software. Unfortunately this inability was not confined to complicatedestimations, e.g. potential monetary savings from a greater stability of OSS, but alsoto more simple questions like license fee savings or hardware cost savings.

Importance of different criteria for usage decisions asked

We decided therefore to ask the surveyed establishments for an indication of how im-portant each item in a list of potential criteria was for their last decisions to use OSsoftware instead of commercial software. While the answers to this question do notdirectly translate into monetary value, they do give an indication about the impor-tance of each specific feature of OS software (e.g. no license fees, modifiable sourcecode). The criteria investigated are the following:

❑ Open and/or modifiable source code❑ Lower or no licence fees❑ Better price-to-performance ratio❑ Higher performance❑ Higher stability❑ Better protection against unauthorised access❑ Better functionality❑ Higher number of potential applications❑ Open Source software was already integrated in another product the establish-

ment had acquired❑ Hardware cost savings❑ Cost savings regarding installation, integration and customisation to compa-

ny needs❑ Cost savings regarding daily operations, administration and support❑ Cost savings regarding training and introduction of users❑ Recommendation of the establishment’s IT service provider❑ Existing solutions, know-how and/or experiences in the establishment regard-

ing the use of Open Source software for the specific purpose.List of criteria derived from OSS literature and TCO model

This list of criteria was derived from an extensive review of the literature about OSSuse and its potential advantages. In addition common concepts for the evaluation ofIT investments, especially the concept of “Total Cost of Ownership” went into thecreation of this list. This concept boils down to the idea that not only direct softwarelicense cost are to be taken into account for a decision but also indirect costs like those


5.1 Benefits from using Open Source software34

for educating users, for support, for integration into the IT infrastructure, etc. Takingall these aspects into account might lead to the result that a software with low licensecost suddenly becomes expensive and vice versa.

Criteria tested on fourusage areas of OSS

We tested the importance of these criteria for the following four usage areas for OSS:

❑ Open Source software as server operating system (e.g. Linux or Free/OpenBSD),

❑ Open Source software for databases (e.g. MysQL, PostgreSQL, Interbase,SAP-DB),

❑ Open Source software on desktop or client computers (e.g. Linux, KDE,Gnome, Mozilla, StarOffice/OpenOffice), as well as

❑ Open Source software in connection with creating or operating websites (e.g.,Apache, PHP, Perl, Python, Squid or Open Source content management sys-tems).

Distinction between areasbecause of different

software as well as differentinvestment calculations

Distinguishing between these aspects seemed important as different software comesto use in all these areas. Therefore, even if the same software decision is considered(e.g. Linux vs. Windows), a total cost of ownership calculation (e.g. for server vs. cli-ent use) might come to different results.

Unweighted results relateto those establishments

using OSS software in area

Contrary to the results reported in the previous chapter, the numbers reported hereare unweighted. I.e., all numbers relate to the sum of those entities in the sample thatindicated they were using OSS within this area or were planning to do so within thenext year.

5.1 Benefits by usage area

5.1.1 OSS used for server operating systems

56% of OSS users employOSS as server operating

system

Of those 395 establishments surveyed, 220 indicated that they are either using OpenSource software for server operating systems or are planning to do so. That corre-sponds to an unweighted share of almost 56% of all OSS users.

Importance of different server operating systems

Apart from Linux, the best-known Open Source operating system, also some otheroperating systems exist that are Open Source. Free/Open BSD is probably the secondbest in popularity. Those people surveyed that indicated they were using another OSserver operating system, often gave Linux distribution brand names, e.g. Redhat, Suseor Mandrake.

Linux by far the mostimportant server operating

system

As figure 5–1 indicates, Linux is indeed “the” Open Source server operating system.Overall 172 establishments or 78% of those that already use OSS on their servers orplan to do so employ Linux. Another 12% are planning to do so within the next year.Taking into account that some of those surveyed did not know that Redhat etc. areLinux distributions, the real percentage of Linux users is even slightly higher. Com-pared to all establishments using any kind of OS software, the share of Linux on serv-ers is remarkable 44%.



Figure 5–1Usage of different OS server operating systems

Base: Establishments currently using or planning to use OSS as server operating system.

Criteria for the decision in favour of Open Source software

Product characteristics are most important decision criteria: OSS is perceived to have better stability…

Figure 5–2 summarises the importance of criteria for the server operating system de-cisions. Quite clearly, product characteristics that are indispensable for a server oper-ating system come highest in importance. Higher stability and better accessprotection than proprietary solutions are the most important decision factors in fa-vour of an Open Source server operating system. Stability is the most important partof any server operating system and Unix-based operating systems have a better stabil-ity record than most others. As Linux and BSD are Unix variants, it is not surprisingthat almost 83% considered this to have been an important or very important factorinfluencing their decision in favour of OSS.

Figure 5–2Importance of all criteria for decision in favour of OS server operating system

*

Other

Free/ Open BSD

Linux

0% 20% 40% 60% 80% 100%

11% 1%

9% 2%

78% 12%

currently using planning use within next year

Ber

leco

n Re

sear

ch 2

002

n=220

Integration in acquired product

IT service provider recommendations

Training cost savings

Higher number of potential applications

Hardware cost savings

Existing solutions and know-how

Open and/or modifiable source code

Installation and integration cost savings

Operation and administration cost savings

Better functionality

Higher performance

Better price to performance ratio

Low license fees

Better access protection

Higher stability

0% 20% 40% 60% 80%

3% 16%

6% 18%

6% 21%

11% 29%

12% 29%

12% 35%

12% 33%

16% 37%

16% 44%

19% 47%

20% 53%

27% 44%

32% 39%

40% 35%

44% 39%

very important important

Ber

leco

n Re

sear

ch 2

002

n=220



…and better protection tounauthorised access

But also protection against unauthorised access is an important feature of any serveroperating system. Obviously the security features of Linux and BSD are convincingfor most decision makers. Almost 75% indicated that better protection against unau-thorised access than in proprietary software has been an important or very importantfactor for their last decisions in favour of an Open Source server operating system.

70% value license costsavings and better price-to-

performance ratio

Next comes the low or not existing direct license fee. 71% indicated that on averagethese have been an important or very important criterion for their decision in favourof OSS (mostly Linux). Closely related is the price-to-performance ratio. Also 71%indicated that a better price-to-performance ratio has been at least an important rea-son for choosing OSS on their servers in the past.

Higher performance andbetter functionality

important, but less oftenvery important

Closely following in importance are again product characteristics. 73% had the feel-ing that their OS server operating system was simply better than a commercial alter-native by providing better performance. 66% considered the better functionality oftheir Open Source server operating system to have been an important decision crite-rion. However, compared to the previous two product characteristics, higher per-formance and better functionality were mostly rated important and much less oftenvery important.

Indirect monetary benefitsof lesser importance than

direct savings

Indirect monetary benefits in the form of cost savings in daily operations, adminis-tration and support as well as in installation, integration and customisation to com-pany needs were considered to have been at least important for 60% and 53%,respectively, of those that use Open Source server operating systems. Although we donot know (and probably most establishments do not either), how large these cost sav-ings were, they have nevertheless to be considered an important aspect.

Compared to the direct license cost savings, however, these indirect monetary effectsfrom using Open Source server operating systems seem to be of lesser size. Otherwisethe establishments would have considered them to be more important. Cost savingsin the installation process or in operating the servers are only considered by half asmany establishments to have been very important as the direct license cost savings are.Hardware cost savings have been even less important. The fraction considering themto have been very important is less than a third of that for direct license cost savings.

Visibility and modifiabilityof source code not major

criterion

The major characteristic of Open Source software, free access to and the possibilityto modify the source code has not been a major factor influencing the decisions of oursurvey respondents. Only for 45% it has been an important or very important deci-sion criterion in favour of their OS server operating system. More important havebeen direct and indirect cost savings from Open Source server operating systems aswell as their product characteristics.

Existing solutions, know-how, and experience atleast important for 47%

Existing solutions, know-how, and experience with OS server operating systems havebeen at least an important decision factor for 47% of Open Source server operatingsystem users. This implies that once companies or public institutions start using OSsoftware, there is a high probability that they will continue to use it. This correspondsquite well to the result from section 4.2 above, which showed that OSS users considerit to be a more important part of their IT infrastructure if they use it in regular oper-ations.

Hardware cost savings oflower importance …

While Linux and other Open Source server operating systems are often propagatedwith the argument that they need less expensive hardware than other variants of Unixor that existing hardware can be used, only about 40% of respondents indicated thatthis has been an important factor for their last decisions in favour of an Open Sourceoperating system. This outcome might be due to Linux competing more with Win-dows (both are using the same hardware) than with other, larger Unix variants.



… as is higher number of potential applications

Also not very important was a higher number of potential applications. Only 40% ofrespondents indicated that this has been an important decision criterion for their de-cision in favour of an OS server operating system.

Last in importance were training cost savings, recommendations and integration in other product

The last three criteria are training cost savings, IT service provider recommendationsand the integration of OSS in another product. For each of these far less than 10%of respondents indicated that they have been very important. Integration of OSS inanother product could have been an important factor, as sometimes the decision foran operating system is not made by the establishment itself that will be using the soft-ware, but by a software manufacturer that already has integrated the operating systeminto his product. This is the case, for example, for appliances, combinations of single-purpose hard- and software that are often offered as communication servers or fire-walls, among other things. It is – what many might not know – also the case with theMacintosh operating system MacOS X, which is based on a BSD-based Open SourceUnix called Darwin. For most of the respondents, though, this was not an importantfactor during the last one to two years. Only 19% considered it to have been impor-tant or very important.

5.1.2 OSS used for databases

Databases are important element in IT infrastructure

Databases are a very important element within the IT infrastructure of every compa-ny or public institution. While several software applications have databases included,there are many circumstances where the establishments have installed stand-alone da-tabases hosting a significant part of its data. This is most often the case, when severalapplications are supposed to access the same data (e.g. customer, product or sales da-ta). It is also the case when IT departments combine so-called best-of breed softwareinto a software solution meeting their requirements. (Examples for such a packagebased on Open Source software are so-called LAMP systems consisting of Linux,Apache, MySQL and PHP.) In these cases companies and public institutions make adeliberate decision about which database product to choose.

Primarily stand-alone databases covered

It is primarily such deliberate database usage decisions that are covered in our survey.Other decisions, where databases are part of a pre-fabricated package, are only includ-ed if the survey respondent knows about the underlying database and if the latter hasinfluenced her decision somehow.

Importance of different Open Source databases

MySQL most used Open Source database

Overall 167 of the surveyed 395 establishments (42%) are using some sort of OpenSource database or are at least planning to do so within the next year. As most wouldhave expected, MySQL is the most-used Open Source database. 71% of the OS da-tabase users either currently use MySQL or are planning to do so. Second are Post-greSQL and SAP-DB with 14%. While PostgreSQL is fairly well-known as OSdatabase, the popularity of SAP-DB is astonishing. This database product is rathercomplex and thus especially suited as basis for SAP and other enterprise applica-tions.14

14. We cannot state with certainty that all respondents really understood the difference betweenSAP, the enterprise resource management solution, and SAP-DB, the SAP database. This mighthave biased the outcome upwards, as SAP is quite widespread, especially among German enter-prises.



Figure 5–3Usage of different OS

databases

Base: Establishments currently using or planning to use OSS as database.

The category of “other” database products has to be interpreted with care. It includesOpen Source databases like HyperSQL, but also database products like IBM’s DB2and Oracle, which are not Open Source software. They are, however, available in var-iants for Linux, which might have led some respondents to confuse the exact meaningof “Open Source database”.

Criteria for the decision in favour of Open Source databases

Ranking of criteriaresembles choice of OSSserver operating system

Figure 5–4 shows in some aspects a rather similar picture to figure 5–2 for OSS serveroperating systems. Most important decision criteria in favour of OSS databases havebeen product features and license cost savings. Indirect cost savings and modifiabilityof the source code have been of lesser importance. Thus, also the database usage de-cision results support the hypothesis that modifiability of the source code is not a ma-jor direct benefit of Open source software for enterprises. It might be an indirectbenefit though, in that an OS code makes a continuous improvement of the softwaremore likely, most important with respect to stability and security. Companies mightmake decisions in favour of OSS since they know that this software went through thisfireproofing process and will continue to do so.

Protection againstunauthorised access of

utmost importance

As databases contain crucial data of each institution, security in the sense of protec-tion against unauthorised access is of utmost importance. Consequently, better accessprotection was the most important reason why establishments decided in favour ofOSS databases. 34% of respondents considered it to have been very important andfor further 37% it was an important factor for their decision in favour of an OSS da-tabase.

Also stability crucial featurefor database

Also stability is one of the main requirements for a database system, especially if thedatabase is a crucial part of a company’s IT infrastructure. As figure 5–4 shows, sta-bility is obviously considered to be one of the main advantages of Open Source data-bases, even more important than the low price. 77% of the respondents stated that ahigher stability has been an important reason why they decided in favour of an OpenSource database. For 31% it has even been a very important reason. While these val-ues are absolutely high, they are lower than in the server operating system decisions.

*

Other

Interbase

PostgreSQL

SAP-DB

MySQL

0% 20% 40% 60% 80% 100%

10% 3%

4% 2%

11% 3%

12% 2%

54% 17%


Ber

leco

n Re

sear

ch 2

002

n=167



Figure 5–4Importance of all criteria for decision in favour of OS database

License fee savings and better price-to-performance ratio equally relevant

Just as for the server operating system, the low or even zero license fees for the OpenSource database have been a major element in the establishments’ decision in favourof an Open Source database. 71% considered it to have been at least important. Theinfluence of the price-to-performance ratio on the database decision was regarded asalmost equally important. Again 71% considered a better price-to-performance-ratioto have been at least important in their decision. 21% considered it very important,which is somewhat less than the price tag alone that was considered to have been im-portant for 28%. Just like in the server operating system decision, this can be inter-preted in a way that getting the software (almost) for free was a very importantdecision criterion, but that the users consider proprietary software to deliver a betterprice-to-performance ratio at least in some cases.

Higher performance has been relevant for 68%

This is supported by the relevance attached to higher performance, again an essentialelement of databases. In many cases, the Open Source database obviously was simplythe better software, as figure 5–4 shows. 69% of the respondents stated that a higherperformance has been at least important for their last decisions in favour of OpenSource database software.

Indirect cost savings due to easier installation and administration

Indirect cost savings are next in relevance. 62% considered installation and integra-tion cost savings to have been important for their decision in favour of Open Sourcedatabases. This value is also considerably higher than for server operating systems.Obviously, much of proprietary database software is considered to be much more dif-ficult to install and to integrate than the available Open Source databases. Almostequally important have been expected savings regarding daily operations, administra-tion and support. 60% considered them to have been an important or very importantcriterion for their decision. This fraction is about as large as for Open Source serveroperating systems.

Better functionality important criterion for 65%

While stability and security are typically perceived as the major advantages of OpenSource software, opinions differ on whether Open Source software provides better











Higher performance


Low license fees

Higher stability


0% 20% 40% 60% 80%

7% 33%

7% 26%

7% 34%

9% 37%

10% 31%

11% 34%

12% 38%

15% 50%

16% 46%

17% 43%

20% 49%

21% 50%

28% 43%

31% 46%

34% 37%


Ber

leco

n Re

sear

ch 2

002

n=167



functionality. Also our surveyed establishments that have made decisions in favour ofOpen Source databases emphasise it to a lesser extent as a decision criterion than sta-bility and security. Nevertheless, for 65% better functionality of Open Source soft-ware has been an important criterion in favour of an OSS database, for 15% even avery important one.

Openness and modifiabilityof lesser importance

Also for OS databases the major characteristic of all OS software – open and modifi-able source code – proves to be of relatively low importance. It was considered to havebeen an important element in the last decisions by almost 50% of the respondents –a slightly higher percentage as with server operating systems. However, it has onlybeen very important for 12%.

Existing know-how hasbeen important for 45%

Existing solutions, know-how and experience have influenced the decision in favourof Open Source databases for 45% of the respondents in an important or very impor-tant way. This is about the same importance as with server operating systems.

Hardware and training costsavings of low importance

Hardware and training cost savings are also for databases of lesser importance thanother forms of direct and indirect cost savings. Both are only considered by 41% ofOS database users to have been an important decision criterion. Reason for the rela-tively low importance of the first is probably that OS databases as well as many pro-prietary ones are available for a large variety of hardware, so that both can be chosenindependently. Reason for the low importance of training cost savings might be thatall major databases use dialects of SQL, the standardised query language.

Importance of highernumber of applicationsreflects potential use of

databases as basic elementwithin other applications

46% consider a higher number of potential applications to have been an importantcriterion in favour of Open Source databases in their establishments. That number isconsiderably higher than for server operating systems. This can either indicate thatserver operating systems do not need a high number of applications or that there aresimply more available for Open Source databases than for Open Source server oper-ating systems. Most likely the second is the more important argument, as the next cri-terion shows.

40% of the respondents stated that the existing integration in another acquired prod-uct was an important criterion in favour of an Open Source database. For server op-erating systems this value was only 18%. Many Internet-technology basedapplications (e.g., groupware applications, content management systems) have OpenSource databases included, which might explain the outcome to some extent.

IT service providerrecommendations least

important

IT service provider recommendations are again the least important motivation be-hind a decision for Open Source software. Only 32% indicated that they have beenan important factor influencing their decision in favour of OS database software, andonly for 7% they have been very important. Nevertheless, this percentage is higherthan for server operating systems.

5.1.3 OSS used on desktop computers

Desktop software related toclient operating systems…

Open source software on desktop computers can come in different variants. First ofall, OSS might be the client or desktop computer operating system, e.g. Linux. Sec-ondly, it might be a certain part of software that is usually considered part of the op-erating system but developed by different developers than the operating system core.The KDE or Gnome desktops are examples for such software. Most often they areused in combination with Linux.

… or as standaloneapplications

And thirdly there are applications. Quite contrary to Linux or KDE they do not formpart of the basic software on a computer, but are application programs for the end



user. The web browser Mozilla belongs into this group as does the office packageOpenOffice/StarOffice. These application programs are often available for a varietyof platforms. Their popularity is therefore not limited to niche operating systems.

Importance of different Open Source components for desktop computers

OSS on desktops not very widespread

Altogether the use of Open Source software on client or desktop computers is notvery widespread. Only about 20% of those establishment that use OSS have someform of OSS installed on their desktops. Of all four application areas asked about,this is by far the lowest value. It reconfirms the common perception that Open Sourcesoftware is mostly used on servers and in other hidden parts of the IT infrastructure.

Most widespread OS desktop software is Linux

Figure 5–5 shows the different sorts of Open Source software used on client or desk-top computers. Of the 80 establishments 68% use Linux on desktop computers orplan to do so. Thus, Linux is according to our survey the most important desktopsoftware using the Open Source model. Next in popularity is StarOffice/OpenOffice,which is currently or within the next year used by 48% of the respondents, followedby the two desktop systems KDE and Gnome. Mozilla, the web browser, will be usedby 28% of the respondents if their plans materialise.

Decision criteria reflect decision for operating system as well as for application

Thus, the answers to the decision criteria analysed next will be a mixture of answersfor an operating system, a desktop, and application software. While these are indeeddifferent sorts of software, they all have in common that they are installed on a clientcomputer meaning that there are typically many people using this software (we ex-cluded use in exceptional cases) and that these users are typically not computer ex-perts.

Figure 5–5Usage of different OS desktop software

Base: Establishments currently using or planning to use OSS on client or desktop systems.

*

Other

Mozilla

Gnome

KDE

StarOffice/ OpenOffice

Linux

0% 20% 40% 60% 80% 100%

8% 3%

23% 5%

23% 4%

31% 8%

33% 15%

49% 19%

currently using planning use within next yearB

erle

con

Rese

arch

200

2

n=80



Criteria for the decision in favour of Open Source software on desktops

Figure 5–6Importance of all criteria

for decision in favour of OSdesktop software

Results again similar toother sorts of Open Source

software …

Figure 5–6 shows the importance of all criteria for acquiring Open Source desktopsoftware. The chart shows a quite similar picture to the previous charts. The productfeatures stability and security have been the main reasons why the establishments havedecided to use OSS on their desktops, followed by direct cost savings due to low orzero license fees and subsequently followed by indirect cost savings.

… with some variations inorder

Compared to the previously analysed areas, the order is only slightly changed. Havingexisting solutions, know-how and or experiences was a less important factor for thedecision in favour of OS desktop software than it was for the decision in favour of anOS server operating system or an OS database. Also better functionality is consider-ably lower ranked than for server operating systems

Security most importantdecision criterion

Security also is the most important decision criterion for desktop software, as the sur-vey results show. 41% considered better protection of OSS desktop software a veryimportant criterion for their decision, altogether 71% considered it to have been atleast important for their decision. Obviously most users consider these possibilities tobe better for Open Source software than for proprietary alternatives.

Higher stability ofrelevance for 80%

Next in importance is higher stability. If the assessments as important and very im-portant are combined, it is even the major reason why those establishments surveyeddecided in favour of Open Source software on their client desktops. 80% indicatedthat it has at least been important, for 35% it has even been a very important criteri-on. These values were only higher for Open Source server operating systems. Obvi-ously, many IT decision makers are not satisfied with the unstable client operatingsystems and software applications they otherwise have available.

Low license fees moreimportant than in previous

usage areas

Also the importance of low license fees is considered to have been larger than for theprevious two usage areas. For 78% low or zero license fees have been at least an im-portant reason for using Open Source software on desktop computers. This was to be












Higher performance

Low license fees

Higher stability


0% 20% 40% 60% 80%

4% 29%

6% 20%

11% 48%

15% 30%

15% 39%

18% 38%

19% 41%

20% 43%

21% 45%

21% 41%

24% 53%

25% 40%

34% 44%

35% 45%

41% 30%


Ber

leco

n Re

sear

ch 2

002

n=80



expected, as license fees can become a significant cost factor if licenses for many com-puters have to be bought. Typically this is more likely to be the case for clients thanfor servers.

Higher performance not as important as direct cost savings

Closely related is again the question for the importance of a better price-to-perform-ance ratio. The results are very similar to those for the importance of low license fees.However, the percentage of respondents considering it to have played a very impor-tant role is lower than for the low license fee. This outcome fits with the answersabout the importance of higher performance. For 65% higher performance has beenat least an important reason for their acquisition of Open Source desktop software.For 25% this reason was even very important. These are both lower values than forthe license fee criterion.

Indirect cost savings for installation and administration highest across areas

Indirect cost savings are given a relatively high importance from users of OS on desk-tops. While 66% consider savings on installation and integration costs to have beenat least important for their decision, 63% come to the same conclusion regarding op-erating and administration cost savings. These are the highest values across all fourusage areas. Obviously proprietary software is regarded as being suboptimal with re-spect to costs for installation and administration.

Better functionality and open source code equally important

Close together are the assessments of better functionality and modifiable source code.Better functionality of OS desktop software was an important or very important rea-son for their decision for 63% of those establishments surveyed. As in the other areasanalysed, this value is somewhat below those for the lower license fees or the higherstability and better access protection. In comparison, 60% of those using OSS ondesktops considered the openness and modifiability of software to have been at leastan important decision criterion and 19% even consider it to have been very impor-tant. These are larger numbers than for all other usage areas, indicating that the wishto modify software is most pronounced for desktop software.

Also hardware cost savings and higher number of potential applications with highest importance across areas

Expected hardware cost savings were of medium importance. For 56% of those usingOS desktop software, these have been at least an important decision factor. Again,this value is higher than for all other usage areas. Being able to use smaller computersor use older computers for a longer time can lead to significant cost savings, as thesavings on each single computer have to be multiplied by the number of computersin a company.

Again somewhat less important was the higher number of potential applications.54% of the users of OS software on desktops said that this influenced their decisionin favour of OSS at least in an important way. This is, however, a higher value thanfor all other areas. Additional applications are obviously more important for desktopusage than for server operating systems or databases. The latter are often acquired fora specific purpose.

Training cost savings with lowest importance of indirect cost savings

Training cost savings have only been an important reason for deciding in favour ofOS desktop software for 45%. This value is lowest of all potential indirect benefitsfro using OS software, as it is in the other areas. OS software obviously does not leadto substantial training cost savings.

Existing solutions and know-how important but not very much

Existing solutions and know-how are considered by fewer respondents to have beenvery important but by more to have been important. Overall 59% of those actuallyusing OSS on desktops have decided to some extent so because they had existing so-lutions, know-how or experience that would facilitate the use of this software in theestablishment.



Service providerrecommendations low on

the scale

Recommendations of the IT service provider are very low on the scale of factors fa-vouring the use of OS software on desktops. Only 26% of respondents indicated thatthey have been at least an important factor in their last decisions in favour of OSSsoftware and only 6% indicated that they have been a very important factor. Thesevalues are comparable to those for server operating systems.

Integration in otherproducts seldom the case

for desktop software

Integration in another acquired product is seldom the case for desktop Open Sourcesoftware. Therefore only 4% said that this has influenced their decision in a very im-portant way. However, 33% indicate that it has influenced their decision at least inan important way. This is probably due to certain components (KDE or Gnome forexample) that are typically bundled with the basic operating system.

5.1.4 OSS used for creating and operating websites

Many popular OS softwareapplications in this field

Apart from the operating system Linux, that application field with most popular ex-amples for Open Source applications is the creation and operation of websites. Webserver projects like Apache, script languages like PHP or Perl as well as special appli-cations like the caching software Squid are much used in this area and fairly well-known. Even more complex applications like content management systems on OpenSource basis are being developed. Thus the answers to this usage area cover again avariety of different applications and application areas.

39% of the OSS usingestablishments employ

OSS software in this area

39% of the OSS using establishments employ Open Source software in connectionwith the creation, maintenance and operation of websites. That is a slightly lower per-centage than for the usage of OS databases but significantly more important than theuse of Open Source software as desktop operating system.

Importance of different Open Source components for websites

Figure 5–7Usage of different OS

software in connection withwebsites

Base: Establishments currently using or planning to use OSS as server operating system.

*

Other

Python

OS Content Management

Squid

PHP

Perl

Apache

0% 20% 40% 60% 80% 100%

8% 2%

8% 1%

10% 6%

26% 3%

32% 8%

48% 6%

72% 10%


Ber

leco

n Re

sear

ch 2

002

n=155



Apache most popular in this area, second is Perl

Quite many Open Source programs exist in this area, leading to a rather diverse usageof this sort of software. First of all, there is the possibility of using Open Source webservers like Apache. Indeed Apache turned out to be the most popular application inthis field. 72% of those using any sort of OS software in connection with websitesemploy Apache. Second in popularity is Perl, used by 48% of OS users. PHP comesthird with 32%.

Combining these results with the popularity of Linux as server operating system andMySQL as Open Source database, the acronym LAMP should better stand for Linux,Apache, MySQL, Perl instead of PHP at the end. Nevertheless, to a large extend bothPerl and PHP can be used for the same purpose and in similar ways for the operationof websites.

Squid used by 26% of respondents in this area

Squid, the Open Source caching software, is typically used in larger Intranets to re-duce the amount of traffic on an establishment’s external Internet connection. It istherefore more used to manage retrieval of websites than the operation of an estab-lishment’s own websites. 26% of those respondents using any sort of OS software inconnection with websites employ Squid.

Tomcat most popular among “Others”

The programming language Python is used by only 8% of respondents in this area.It is thus much less popular than Perl with which it has several features in common.Of similar popularity are other sorts of Open Source software. Apart from some peo-ple mentioning Linux, PHP-Nuke – an OS content management system – was men-tioned a few times in the category “Others”. Most often indicated was the use ofTomcat. Tomcat is the servlet container that is used in the official reference imple-mentation for the Java Servlet and JavaServer Pages technologies and is by now partof the Apache project.

Open Source CMSs not very widespread

Open Source content management systems are used by roughly 10% of the establish-ments in this group.

Criteria for the decision in favour of Open Source software for websites

Similar picture than for other usage areas

Figure 5–8 combines the importance of all decision criteria in favour of OSS in con-nection with websites. Despite some slight differences in the details, the basic orderof importance is the same as for the other areas. The most important criteria in favourof Open Source software are software characteristics like higher stability and betteraccess protection. Thus, OSS is perceived by most users as being simply better thancommercial alternatives.

Direct cost savings more important than indirect cost savings

Low or zero license fees are next in order of importance, very close to higher perform-ance, better price-to-performance ratio and better functionality. This group consistsof direct monetary benefits from using this software as well as of benefits from betterquality. Indirect cost savings together with the openness and modifiability of thesource code make up the next category. As in all other categories, cost savings in in-tegration and installation as well as in daily operations are more important than hard-ware cost savings. The importance of an open source code is right in between theseindirect benefits from choosing Open Source software.

Higher stability primary reason for choosing OSS

Higher stability of Open Source software is the primary reason for choosing OSS inconnection with websites. 83% state that it has been at least important and for 48%it has even been a very important factor. Of all criteria and all domains these valuesare the highest and thus show that OSS obviously fits the requirements for website-related software best.



Figure 5–8Importance of all criteriafor decision in favour ofOSS in connection with

websites

Better access protectionsecond most important

While access protection is also an important issue for software used in connectionwith websites, it was of lesser importance in choosing OSS than several other criteria(stability, price-performance ratio) were. 74% stated that the better access protectionprovided by OSS had been at least an important reason for choosing it. A reason forthis outcome might have been the diverse nature of software discussed in this field.PHP, Perl and Python, for example, are software as well as programming languages.For programming languages protection against unauthorised access is not an impor-tant feature, though.

Low license fees, higherperformance, better

functionality and betterprice-to-performance ratio

close together

Quite closely together are low license fees, higher performance, better functionalityand – as consequence – a better price-to-performance ratio. For all of these criteria,between 70% and 77% indicated that they have been at least an important reason fortheir decision in favour of OSS. For 29-32% they have even been a very importantreason. Even slightly more important than the price alone was the better price-to-per-formance ratio of Open Source software in this area. We would expect that this resultis to some extent influenced by Apache, which is often said to provide better perform-ance than commercially available web servers.

For 56% openness andmodifiability of source code

important

56% stated that the openness and modifiability of source code has been at least animportant reason for their choice of Open Source software in this area. These valuesare of the same size as for the choice of desktop software and higher than for serveroperating system and databases.

Indirect cost savings As in the other usage domains, indirect cost savings are less important than productcharacteristics and direct cost savings. 55% indicate that savings in daily operationsand administration have been important or very important for their decision. Like-wise, 53% indicate that installation and integration cost savings has been importantor very important. This is also lower than in most other domains. Of slightly less im-portance are existing solutions, know-how and experiences. For 50% of respondentsthey have been at least an important reason for choosing OS software.












Higher performance

Low license fees


Higher stability

0% 20% 40% 60% 80%

6% 19%

6% 21%

10% 23%

10% 38%

11% 27%

12% 38%

16% 37%

19% 37%

20% 35%

30% 41%

30% 47%

32% 41%

32% 41%

44% 30%

48% 35%


Ber

leco

n Re

sear

ch 2

002

n=155



Hardware cost savings, higher number of potential applications and training cost savings group with lower importance

A group with even lower importance is made up of hardware cost savings, a highernumber of potential applications and training cost savings. Only 38% stated thathardware cost savings have influenced their decision in an important or very impor-tant way favourably. For only 11% they have constituted a very important motiva-tion. The other two criteria yield similar values.

Finally, IT service provider recommendations as well as the integration in acquiredproducts are again on the last positions. Only for 27% of the respondents have rec-ommendations by IT service provided an important incentive for choosing OpenSource software. This criterion was therefore equally unimportant as the inclusion inanother product. Only 25% state that this had been an important or very importantfactor for their decision. This value is second-lowest after server operating systems.

5.1.5 Summary of OSS benefits for companies and public institutions

Consistency across application areas most striking outcome

The most striking result of evaluating the OSS selection criteria was their consistencyover all four areas of application. No matter, whether one considers the use of OSSas server operating system, as database, as desktop software or in connection withwebsites, the basic picture remains the same. This basic picture is characterised byfour observations:

1. Higher stability and better access protection most important

Higher stability and better protection against unauthorised access are the most im-portant reasons why the surveyed establishments have made decisions in favour ofopen source software and against proprietary competitors. Higher performance is alsoan important reason why the respondents decided in favour of Open Source software.Thus, most beneficial to professional users of Open Source software are those specificproduct characteristics that are often said to be a direct consequence of the specificdevelopment process of Open Source software.

2. Low or zero license fees come second in importance

Low or zero license fees come second in importance. The fact that Open Source soft-ware is typically distributed without charge or only for a nominal fee is an importantfurther reason why companies and public institutions use Open Source software. Thedirect benefits in form of license fee savings are thus more important than indirectcost savings.

3. Installation and administration cost savings come third

Indirect cost savings from using Open Source software come third in importance. Butthis applies mainly to two sorts of cost savings, those regarding installation, integra-tion and customisation to company needs and those regarding daily operations, ad-ministration and support. Other forms of potential cost savings, i.e. relating tohardware or user training turned out to be of lesser importance.

4. Open and modifiable source code comes only fourth for professional users

Only fourth comes the open and modifiable source code, the characteristic that de-fines Open Source software. According to the survey results, this openness is not amajor reason why companies and public institutions use Open Source software. Itmight however, be an indirect criterion in the sense that companies believe that soft-ware developed under an OS model is better, e.g. more stable and more secure. Sucha claim for Open Source software is more believable since everybody can test it, as thesource code is open, even though many people do it.15

15. There is a parallel to science, where much of the authority comes from the possibility to checkresults and repeat experiments, but in most disciplines this is not done for many results.



5.2 Establishment characteristics and OSS decisions

Influence of establishmentcharacteristics on

importance of decisioncriteria

While the importance of those aspects discussed in the previous section for the OSSusage decision speak for themselves, it would also be interesting to know whetherthere are certain establishment characteristics that influence these decisions. Do, forexample, small establishments attach a greater importance to direct cost savings dueto zero license fees? Or does the public sector consider certain aspects as significantlymore important than private sector establishments?

Series of 60 regressionswith 12 independent

variables

To obtain answers to these questions, a series of 60 regression analyses has been con-ducted with the answer to each of the items in the four usage areas as dependent var-iables and a set of different explanatory variables:

❑ The country where the establishments is located: UK, Sweden or Germany,❑ The sector to which it belongs: private sector with low medium or high IT in-

tensity or the public sector,❑ Whether the establishment is small, i.e. belongs to the size class of 100-499

employees,❑ The share of IT personnel in the establishment,❑ The relation of computers within the establishment to employees,❑ The importance of OSS for the establishments’ IT infrastructure,❑ The degree to which an establishment uses OSS because it wants to be inde-

pendent from the pricing and licensing policies of big software companies,❑ The degree to which an establishment uses OSS because it wants to support

the OS community by using Open Source software,❑ The degree to which an establishment uses OSS since it regards OS specialists

to be more easily availably on the labour market,❑ The degree to which an establishments uses OSS because doing so is company

policy,❑ The degree to which an establishment lets its developers work on OSS projects

on company time, and❑ The degree to which an establishment is working together with OSS service

companies in order to support the development of Open Source software.Discussion only of most

prominent effectsTables 5–1 to 5–4 contain the outcome of these regressions in short form.16 Thoseeffects that were found to be statistically significant are denoted by +, - or by the di-rection they work. A + indicates positive correlation between the variable and the im-portance of a specific criterion, a – a negative correlation. A > or < states that theparameter value for establishments denoted on the left (e.g. German or High IT in-tensity establishments) is significantly higher or lower than that for establishments de-noted on the right. Note that a higher parameter value means that establishmentswith this characteristic attach a lower importance to this item.

Only few consistent andstable relationships

between explanatoryvariables and importance of

different criteria

There are only very few effects consistent across the four usage areas for Open Sourcesoftware. Most are only observable for one or two combinations of usage area and var-iable. For example, small establishments have stated more strongly than large compa-nies that better protection against unauthorised access has influenced their decisionin favour of Open Source for website related software. There is no such significanteffect for the other usage areas. We will therefore discuss in the remainder of this sec-tion only those effects that have been found to be prevalent in at least three OSS usageareas. If some relationships are strong enough for statements like “For small establish-ments it is more important that …”, than these.

16. The full regression results are available upon request.



UK establishments considered hardware cost savings and recommendations to have been more important

Most of these few relationships concern the regional affiliation of establishments.First of all, the statistical analyses show a couple of occasions, where UK establish-ments considered some items to have been more important than Swedish and Ger-man establishments did. This was the case for the importance of hardware costsavings as well as the importance of IT service provider recommendations.

Integration of OSS in other products less important in Germany

On the other hand German establishments considered the integration of OSS insome other acquired product to have been significantly less important than theircounterparts in the UK and in Sweden did. Open Source software as part of greatersoftware packages thus seems to be more prevalent in these countries.

Availability of source code correlates positively with importance of OSS in IT infrastructure

A quite strong positive relationship can also be observed between the importance at-tached to open and/or modifiable source code and the importance of OSS for an es-tablishment. Thus, establishments that consider OSS to be an important part of theirIT infrastructure also tend to consider the availability of its source code to have beenan important criterion for their decision in favour of OSS. This applies to all usageareas except OSS on desktops, where only few effects are significant due to a lownumber of observations.

Positive correlation between importance of better price-performance-ratio and wish to become independent from large software companies.

A second positive correlation was found to exist between the importance of a betterprice-to-performance ratio as decision criterion and the use of OSS to become inde-pendent from licensing and pricing policies of the big software companies. This effectwas to be expected as it is exactly the zero license fee that makes the software user in-dependent from other licensing schemes. And indeed also in two usage areas a posi-tive correlation between the importance attached to zero or low license fees and thewish to become independent could be observed.

Importance of higher performance goes along with OSS company policy

Finally the third clear correlation exists between the importance attached to higherperformance of OSS software as decision criterion and the degree to which the usageof OSS is company policy. If establishments have such a company policy, they tendto consider higher performance to have been a major criterion for their decisions infavour of using Open Source software.

To summarize these results, there are only very few clear and strong correlations be-tween the explanatory variables and the assessment of benefits. Nevertheless, the ex-planatory power of the regression equations is relatively high and varies mostlybetween 80-90%. The remainder will have to be considered as individual peculiaritiesof using OSS.



Table 5–1 Establishment characteristics influencing choice of OSS as server operating system

Cou

ntry

eff

ects

Smal

lnes

s

Sect

or e

ffec

ts

IT s

hare

PC

sha

re

OSS

impo

rtan

ce in

IT

Pric

ing

and

licen

sein

dpen

denc

e

Supp

ort

OSS

com

mun

ity

IT s

paci

alis

ts b

ette

rav

aila

ble

OSS

is c

ompa

ny p

olic

y

OSS

dev

elop

men

t at

wor

k

Coo

pera

tion

wit

hO

SS s

ervi

ce c

omp.

Open source code + M<P – + +

No license fees + – –

Better price/performance D<S, D<UK + +

Higher performance L<M, H< M + +

Higher stability D<S + +

Better protection D<S +

Better functionality L<M – +

More applications D>S, D>UK

Integrated in other prod. D>S, D>UK + +

Hardware cost savings UK<S, UK<D +

Installation savings

Operation savings +

Traing savings UK<S, UK<D + + – –

Recommendation UK<S, UK<D + – + –

Existing know-how + +

A + indicates positive correlation between the variable and the assessment of importance for a specific criterion, a – a negative correla-tion.A > or < states that the parameter value is significantly lower or higher. Note that a higher parameter value means a lower importance.In column 2 D, UK, and S denote the countries, in column 4 L, M, H, and P denote the three intensities of use as well as the public sector.



Table 5–2 Establishment characteristics influencing choice of OSS for databases

Cou

ntry

eff

ects

Smal

lnes

s

Sect

or e

ffec

ts

IT s

hare

PC

sha

re

OSS

impo

rtan

ce in

IT

Pric

ing

and

licen

sein

dpen

denc

e

Supp

ort

OSS

com

mun

ity

IT s

paci

alis

ts b

ette

rav

aila

ble

OSS

is c

ompa

ny p

olic

y

OSS

dev

elop

men

t at

wor

k

Coo

pera

tion

wit

hO

SS s

ervi

ce c

omp.

Open source code +

No license fees + +

Better price/performance UK<S +

Higher performance L<P,L<M,L<H + + +

Higher stability +

Better protection –

Better functionality –

More applications UK<D –

Integrated in other prod. D>UK, D>S – + –

Hardware cost savings D>UK +

Installation savings UK<S, UK<D +

Operation savings –

Traing savings + –

Recommendation UK<S M<P,M<L,M<H – +

Existing know-how + +

A + indicates positive correlation between the variable and the assessment of importance for a specific criterion, a – a negative correla-tion.A > or < states that the parameter value is significantly lower or higher. Note that a higher parameter value means a lower importance.In column 2 D, UK, and S denote the countries, in column 4 L, M, H, and P denote the three intensities of use as well as the public sec-tor.



Table 5–3 Establishment characteristics influencing choice of OSS on desktop computers

Cou

ntry

eff

ects

Smal

lnes

s

Sect

or e

ffec

ts

IT s

hare

PC

sha

re

OSS

impo

rtan

ce in

IT

Pric

ing

and

licen

sein

dpen

denc

e

Supp

ort

OSS

com

mun

ity

IT s

paci

alis

ts b

ette

rav

aila

ble

OSS

is c

ompa

ny p

olic

y

OSS

dev

elop

men

t at

wor

k

Coo

pera

tion

wit

hO

SS s

ervi

ce c

omp.

Open source code – +

No license fees

Better price/performance L>H –

Higher performance

Higher stability

Better protection


More applications

Integrated in other prod. D>UK, D>S –


Installation savings

Operation savings

Traing savings –

Recommendation UK<S, UK<D L>P, L>M +

Existing know-how H<L




Table 5–4 Establishment characteristics influencing choice of OSS in connection with websites

Cou

ntry

eff

ects

Smal

lnes

s

Sect

or e

ffec

ts

IT s

hare

PC

sha

re

OSS

impo

rtan

ce in

IT

Pric

ing

and

licen

sein

dpen

denc

e

Supp

ort

OSS

com

mun

ity

IT s

paci

alis

ts b

ette

rav

aila

ble

OSS

is c

ompa

ny p

olic

y

OSS

dev

elop

men

t at

wor

k

Coo

pera

tion

wit

hO

SS s

ervi

ce c

omp.

Open source code UK<S M<H + +

No license fees +

Better price/performance +

Higher performance D<S +

Higher stability D<S, D<UK +

Better protection D<S, D<UK + + – + +

Better functionality S>UK, S<D +

More applications +

Integrated in other prod. UK<S, UK<D +

Hardware cost savings UK<S<D L<H +

Installation savings + – – + +

Operation savings S>UK, S>D +

Traing savings UK<S, UK<D +

Recommendation UK<S, UK<D M<L, M<H

Existing know-how UK<S –



556 Survey Questionnaire


Introduction

Good morning/afternoon, my name is _ _ _ _ _ _ _ _ from the _ _ _ _ _ _ _ _ _ in_ _ _ _ _ .

We are conducting a survey on Open Source Software. Our client is Berlecon Re-search, a technology research company in Berlin, Germany. Berlecon Research is wor-king on a research project regarding the professional use of Open Source Software inthe European Union. The project is financed by the European Commission.

I would like to talk to the person responsible for IT decisions and administration inyour company. The person should be able to answer questions about your company’sIT decisions and should have a basic understanding of the technical issues.

Your answers will remain absolutely confidential and the survey results will be pre-sented in an aggregated format only. It will not be possible to draw conclusions aboutyour company from the research results. May I ask you for about 15 minutes of yourtime to answer a few questions?

Question 1-1: Filter

Is your company using Open Source Software, e.g. Linux, Apache, mySQL, or plan-ning to do so within the next year? We are talking about software with source codethat is open, readable and changeable.

Answers: 1=yes, 2=no

If Answer=1 continue, if Answer=2 stop here.

Question 1-2: Warming up

From your point of view, what is the importance of Open Source Software for yourcompany’s IT infrastructure? Is the importance ...

Answers:

1=very high

2=high

3=medium

4=low

5=very low

8=don’t know

0=no answer


6 Survey Questionnaire56

Question 2: IT Areas for Use of Open Source Software

In the following, I will line out specific IT areas for the use of Open Source Software.Please tell me for each area whether your company is currently using Open SourceSoftware or planning to use it within the next year. Please indicate also whether youare using Open Source Software in your regular IT operations or just in unimportantexceptional cases.

Answers:

1=currently using in regular IT operations

2=currently using only in unimportant exceptional cases

3=planning to use within the next year in regular IT operations

4=not using in regular IT operations and not planning to do so within the next year

8=don’t know

0=no answer

❑ Open Source Software for server operating systems, e. g. Linux or Free/Open BSD

❑ Open Source Software for databases, e. g. MySQL, PostgreSQL or Interbase or SAP-DB

❑ Open Source Software on desktop or client computers, e. g. Linux, KDE, Gno-me, Mozilla or StarOffice/Open Office

❑ Open Source Software in connection with creating or operating web sites, e. g. Apache, PHP, Perl, Python, Squid or Open Source Content Management Sy-stems

Transition to Next Question

The following questions will further explore the use of Open Source Software withinthese areas.

Note: Please rotate the question blocks 3-6!

Filter for Question 3:

If the response to Question 2 on the use of Open Source Software for Server Opera-ting Systems has been 1 or 3 go on to Question 3-1. Otherwise go to Question 4.

Transition

Let’s discuss server operating systems.

Question 3-1:

Your company is currently using Open Source Software for server operating systemsor is planning to do so.

In the following, I will name several Open Source operating systems. Please tell mefor each one whether you are currently using it or planning to use it within the nextyear.

Answers:

1=currently using



2=planning to use within the next year

3=not using and not planning to use within the next year

8=don’t know

0=no answer

❑ Linux❑ Free/Open BSD❑ Other, if yes which _ _ _ _ _ _ _ _ _ _ _ _

Question 3-2:

Now I will present to you several criteria that can influence a decision in favour ofOpen Source Software. For your following answers, please refer to all your decisionsin favour of an Open Source server operating system within the last one to two years.

Please tell me, how important each of the following criteria was on average for yourdecision in favour of Open Source and against any proprietary operating system.

Answers: (read after the first criterion)

This criterion was . . .

1=very important

2=important

3=neither nor

4=less important

5=not important

8=don’t know

0=no answer

Criteria rotation

❑ Open and/or modifiable source code❑ Lower or no licence fees❑ Better price to performance ratio❑ Higher performance❑ Higher stability❑ Better protection against unauthorised access❑ Better functionality❑ Higher number of potential applications❑ Open Source server operating system was already integrated in another product

you have acquired❑ Hardware cost savings❑ Cost savings regarding installation, integration and customisation to company

needs❑ Cost savings regarding daily operations, administration and support❑ Cost savings regarding training and introduction of users❑ Recommendation of your IT service provider❑ Existing solutions, know-how and/or experiences in your company regarding

Open Source server operating systems



Transition

Thank you for your answers on Operating Systems.


If the response to Question 2 on the use of Open Source Software for Databases hasbeen 1 or 3 go on to Question 4-1. Otherwise go to Question 5.

Transition

Let’s discuss databases.

Question 4-1:

You are currently using Open Source Software for databases or you are planning todo so.

In the following, I will name several Open Source database products. Please tell mefor each one whether you are currently using it or planning to use it within the nextyear.

Answers:

1=currently using



8=don’t know

0=no answer

❑ MySQL❑ PostgreSQL❑ Interbase❑ Other, if yes which _ _ _ _ _ _ _ _ _ _ _ _

Question 4-2:

Now I will present to you several criteria that can influence a decision in favour ofOpen Source Software. For your following answers, please refer to all your decisionsin favour of an Open Source database product within the last one to two years.

Please tell me, how important each of the following criteria was on average for yourdecision in favour of Open Source and against any proprietary database product.



1=very important

2=important

3=neither nor

4=less important

5=not important

8=don’t know

0=no answer



Criteria rotation

❑ Open and/or modifiable source code❑ Lower or no licence fees❑ Better price to performance ratio❑ Higher performance❑ Higher stability❑ Better protection against unauthorised access❑ Better functionality❑ Higher number of potential applications❑ Open Source database was already integrated in another product you have acqui-

red❑ Hardware cost savings❑ Cost savings regarding installation, integration and customisation to company


Open Source databases

Transition

Thank you for your answers on databases.


If the response to Question 2 on the use of Open Source Software for desktop or cli-ent computers has been 1 or 3 go on to Question 5-1. Otherwise go to Question 6.

Question 5-1:

You are currently using Open Source Software on desktop or client computers or youare planning to do so.

In the following, I will name several kinds of Open Source Software in the desktoparea. Please tell me for each one whether you are currently using it or planning to useit within the next year.

Answers:

1=currently using



8=don’t know

0=no answer

❑ Linux❑ KDE❑ Gnome❑ Mozilla❑ StarOffice/OpenOffice❑ Other, if yes which _ _ _ _ _ _ _ _ _ _ _ _



Question 5-2:

Now I will present to you several criteria that can influence a decision in favour ofOpen Source Software. For your following answers, please refer to all your decisionsin favour of Open Source software on desktop or client computers within the last oneto two years.




1=very important

2=important

3=neither nor

4=less important

5=not important

8=don’t know

0=no answer

Criteria rotation

❑ Open and/or modifiable source code❑ Lower or no licence fees❑ Better price to performance ratio❑ Higher performance❑ Higher stability❑ Better protection against unauthorised access❑ Better functionality❑ Higher number of potential applications❑ Open Source Software was already integrated in another product you have acqui-



Open Source Software on desktop or client computers

Finish Question 5 and Intro to Next Question

Thank you for your answers on the desktop area.


If the response to Question 2 on the use of Open Source Software for creating or ope-rating web sites has been 1 or 3 go on to Question 6-1. Otherwise go to Question 7.



Question 6-1:

You are currently using Open Source Software for creating or operating websites oryou are planning to do so.

In the following, I will name several kinds of Open Source Software for creating oroperating websites. Please tell me for each one whether you are currently using it orplanning to use it within the next year.

Answers:

1=currently using



8=don’t know

0=no answer

❑ Apache❑ PHP❑ Perl❑ Python❑ Squid❑ Open Source Content Management Systems❑ Other, if yes which _ _ _ _ _ _ _ _ _ _ _ _

Question 6-2:

Now I will present to you several criteria that can influence a decision in favour ofOpen Source Software. For your following answers, please refer to all your decisionsin favour of Open Source software for creating or operating web sites within the lastone to two years.




1=very important

2=important

3=neither nor

4=less important

5=not important

8=don’t know

0=no answer

Criteria rotation

❑ Open and/or modifiable source code❑ Lower or no licence fees❑ Better price to performance ratio❑ Higher performance❑ Higher stability



❑ Better protection against unauthorised access❑ Better functionality❑ Higher number of potential applications❑ Open Source Software was already integrated in another product you have acqui-



Open Source Software for creating or operating web sites

Transition

Thank you for your answers on Websites.

Question 7: Open Source Software in General

Now I have some questions on the general use of Open Source Software in your com-pany. They are not related to any specific IT area.

In the following, I will present a number of statements to you. Please, tell me for eachstatement how much it applies to your company. For your answer, you can use thefollowing range:

Answers:

1=totally agree

2=somewhat agree

3=neither nor

4=somewhat disagree

5=totally disagree

8=don’t know

0=no answer

Criteria rotation

❑ We use Open Source Software because we want to be more independent from the pricing and licensing policies of the big software companies.

❑ By using Open Source Software we want to support the Open Source communi-ty.

❑ We use Open Source Software because IT specialists for this kind of software are more easily available on the labour market than specialists for proprietary soft-ware.

❑ We prefer using Open Source Software – that’s part of our company policy.❑ Our software developers are free to work on Open Source projects within their

time at work.❑ We are deliberately working together with Open Source service companies in or-

der to support the development of Open Source software.



Question 8: Company and IT

Finally, I have three questions left concerning your company and your IT.

Question 8-1:

How many IT people does your company employ? (employees in the IT department or employees responsible for IT related tasks)

_ _ _ _ number

8=don’t know

0=no answer

Question 8-2: How many computer users does your company have?

_ _ _ _ number

8=don’t know

0=no answer

Question 8-3:

How many employees does your company have?

_ _ _ _ number

8=don’t know

0=no answer

Final

Thank you very much for your time and your effort. Your evaluations have been veryhelpful for the project. The results of this survey will be published by the end of June2002. They will then also be made publicly available on the web. Look forwww.berlecon.de.


Berlin, July 2002

Lead Analyst:

Dr. Thorsten Wichmann



Motivations and Policy Implications Firms‘ Open Source Activities:

3


B

ERLECON

R

ESEARCH

GmbH


10117 Berlin

Tel.: +49 30 285296-0

Fax: +49 30 285296-29



Acknowledgements:

This work was prepared by Berlecon Research. It is part of the final report for theproject „FLOSS – Free/Libre Open Source Software: Survey and Study“, which wasfinanced under the European Commission‘s IST programme, key action 4 as accom-panying measure (IST-2000-4.1.1).

Disclaimer:





4

V 1.1 - 020905

5


Table of contents

1 Introduction....................................................................................................... 7

2 Large firms‘ Open Source activities..................................................................... 9

3 Motivations for firms‘ Open Source activities ................................................... 17

4 Policy implications of Open Source activities ................................................... 234.1 Open Source activity as standardisation effort ............................................. 234.2 Open Source activity as basic research ......................................................... 27

5 Conclusions...................................................................................................... 31

6 References ........................................................................................................ 33


6

7


1 Introduction

1 Introduction

Creation of Open Source software mostly seen as private activity

The engagement in Open Source (OS) projects and the creation of Open Source soft-ware (OSS) is seen by most commentators and researchers as a private activity. Indi-vidual persons contribute to Open Source projects for private reasons, be it just forfun, in the hope of getting something in return, or because this activity yields privateindirect returns on the labour market.

Current emphasis on reciprocity and labour market motivations

While the assumption of altruistic behaviour dominated the Open Source discussionin earlier years, the current work emphasises more reciprocity or individual labourmarket considerations. E.g., Lerner and Tirole (2002) argue that a programmer cansignal his coding abilities by participating in Open Source projects. This should raisehis expected future wage or give him access to programming jobs, as already Ray-mond (2000, Chapter 5) has pointed out, although he considers the latter as rare andmarginal motivation for most hackers.

Improving job opportunities important reason for participation

While the FLOSS developer survey has shown that developing new skills and sharingtheir knowledge were the primary motivations for participation in Open Sourceprojects, the expected monetary benefits are not negligible: About a third of the sur-veyed developers indicated that improving their job opportunities was a motivationfor their Open Source activity. (Ghosh et al., 2002).

firms‘ deliberate Open Source activities important

Although important for explaining the Open Source phenomenon, this focus on theindividual programmer neglects an important Open Source driver: firms. Part of theOpen Source community consists of individuals employed explicitly for developingOpen Source software. Ghosh et al. (2002) point out that about a third of the sur-veyed developers are being paid directly for developing Open Source software. Thus,their contribution to Open Source projects is the result of firms‘ deliberate decisionsto finance the development of Open Source software. In addition there are several ex-amples of companies that have made available formerly proprietary software as OpenSource software.

Some firms‘ activities are large

The sheer amount of resources devoted by companies to OS development can belarge. IBM alone claims to have spent $1 billion on Linux alone and is also active inseveral other Open Source projects (Wilcox, 2000). Other companies also devoteconsiderable resources to the development of OS software. Due to their size these in-itiatives are important contributions to the development of OS software in total.

firms‘ Open Source activities less well understood than individuals’

Despite this size the companies‘ motivation behind their OS engagement is not aswell understood as the motivation of individual developers. Although discussed inpassing by some authors like Lerner and Tirole (2002) or Schmidt and Schnitzer(2002), much less attention has been devoted to firms‘ Open Source activity than toOpen Source activity of individuals.

Public policy has to consider firms‘ Open Source activities

Also in the context of public policy the firms‘ Open Source activities are of impor-tance. If commercial firms produce Open Source software alongside their proprietarysoftware, a strict distinction between the “commercial world” and the “free world”,as it can sometimes be seen in the public policy debate,

1

might not be sensible. Rather

1 Introduction


8

one would have to ask, whether commercial firms produce the socially optimalamount of Open Source software, just like economics asks whether firms engage in asocially optimal amount of basic research. Even if one comes to the conclusion thatthis amount is sub-optimal, one has to take firms‘ behaviour into account when de-signing policy measures intended to foster the use of Open Source software.

Organisation of this paper

In the remainder of this paper, we will proceed as follows. Section 2 discusses the en-gagement of the world’s 25 largest software companies in OS activities. Section 3groups these activities. This analysis forms the basis for the subsequent work. Section4 discusses the motivation(s) behind the companies‘ engagement in OS developmentusing analogies for similar activities from economic theory. This framework is alsoused to discuss which policy and regulation activities influence the companies‘ behav-iour and which conclusions can be drawn for government bodies and regulatory au-thorities. Section 5 concludes.

1. Often this debate is stylised to (commercial) Microsoft versus (free, open and not commercial)Linux. The many possible shades of grey are assiduously ignored.

9


2 Large firms‘ Open Source activities


Section describes OS activities of large software firms

This section describes the OS activities of globally large software firms. To have a rea-sonable foundation for the subsequent discussion, we investigated the OS develop-ment activities of the world’s 25 largest software companies.

Large firms chosen for two reasons …

We have chosen large firms mainly for two reasons: First of all, we expect the eco-nomic motives behind their OS engagements to be more pronounced in larger thanin smaller companies: As these entities are large and professionally managed, prefer-ences of individuals play a lesser role than they might in small companies. Secondly,these companies typically have sufficient funds available to pursue the strategies theywant to pursue. If there are arguments in favour of an OS engagement, we would ex-pect to see the engagement take place. Smaller companies might in principle come tosimilar conclusions but might refrain from pursuing these strategies due to lack offunds or due to the need to focus on more important things.

… on the basis of Software Magazine’s annual ranking

The basis for this research is the Software Magazine’s 2001 list of the world’s largestsoftware companies,

2

which is published each year and is by now a well-establishedranking. As these 25 companies are typically large, many of them are not softwarepureplays. Some of them are even not in the business of standard software, but areproviders of IT services like consulting and systems integration (e.g. Accenture,PWC) or outsourcing services (EDS).

Web sites and search engines investigated

These companies‘ web sites as well as major search engines have been used to look forsigns of activity in OS development. These activities had to be indicated as being thecompanies‘ activities and not that of individual programmers working there. The out-come therefore ignores “passive” Open Source support such as letting employeeswork on OS projects on company time

3

and thus underestimates the companies‘ totalcommitment of resources to OS projects.

Incentives for firms to make their OS activities known

There are at least two good reasons, why one should be able to find information aboutfirms‘ Open Source activities this way: First of all, OS development takes place pub-licly and typically tries to involve as many contributors as possible. And secondly,Open Source – especially Linux – is a topic with generally positive connotation andhigh growth expectations for the future. Marketing departments therefore have an in-centive to let the world know about a company’s Open Source activities if there areany.

2. http://www.softwaremag.com.3. Some empirical evidence for the importance of the latter is given in Berlecon Research (2002).



10

Tab. 2–1Open Source activities of

the 25 major softwarecompanies

Compamy Open Source activities

IBM Major OS activities.URL: http://www-124.ibm.com/developerworks/oss/

Microsoft No OS activities visible. Microsoft proposes model of “shared source” as alternative to open source.URL: http://www.microsoft.com/licensing/sharedsource/

PriceWaterhouse-Coopers

No OS activities visible. Topic is discussed on web-site from consulting point.

EDS Occasional OS activities.Process data format eXT is proposed to continue as OS after initial deve-lopment by EDS. Also Dynamator, a program for maintaining server pages and developed by an EDS programmer is OS.

Oracle No OS activities visible.

Hewlett-Packard Major OS activities.URL: http://www.hp.com/products1/linux/

Accenture No OS activities visible. Topic is discussed on web-site from consulting point.

Cap Gemini Ernst & Young

No OS activities visible.

Compaq Major OS activities.URL: http://opensource.compaq.com/ URL: http://www.compaq.com/products/software/linux/

Unisys No OS activities visible.

SAP Major OS activities.URL: http://www.sap.com/solutions/technology/linux/URL: http://www.sabdb.org

Computer Associates Major OS activities.Co-founder of Open Source Development Lab.URL: http://www.osdl.org

Hitachi Major OS activities.URL: http://oss.hitachi.co.jp/index-e.html

Sun Microsystems Major OS activities.URL: http://www.sunsource.net

NCR No OS activities visible.

Compuware No OS activities visible. However, development environment shipped with Compurware product OptimalJ is based on the open source Integrated De-velopment Environment (IDE) NetBeans

.

Siebel Systems No OS activities visible.

PeopleSoft No OS activities visible.

SunGard Data SystemsNo OS activities visible.

Fiserv No OS activities visible.

Computer Sciences Corp.

No OS activities visible.

Source: The Top 25 Companies are from Software Magazine‘s 2001 Software 500. Their open source activity has been researched by Berlecon Research in March 2002.

11



Classification into three groups

In a first step the companies were classified into one of three groups: “Major OS ac-tivities”, “Occasional OS activities” and “No activities visible”. Members of the firstgroup are active in several Open Source projects. Typically companies from the firstgroup also have a whole section of their web site devoted to their OS activities ofwhich the URL is provided. The results are compiled in table 2-1.

One third engages in major Open Source activities

Summarizing the results from table 2-1 one can see that almost one third of the 25largest software companies (32%) do engage in major OS development activities.12% (3 companies) have smaller projects and the majority (56%) does not have anyvisible open source projects. At these companies, individuals might be involved in OSactivities, though. Taking into account that OS activities mean that these companiesinvest considerable resources to provide the public with software they cannot (direct-ly) make money off, this number is astonishingly high.

Description of projects in detail to derive common features

To be able to understand the OSS projects and their role within the selection of soft-ware and other products offered by these companies, it is useful to consider the OSactivities of the 9 companies with major OS engagements in more detail. This simpledescription helps to derive common features of the Open Source software compo-nents furthered by these firms.

Closely related are standards and software as well as support and development of OSS

Within the Open Source strategies of these companies, a fewelements are very closelyrelated and cannot always be separated in a satisfactory way. The first such pair isOpen Source software and open standards, as development of standards and referenceimplementations of these standards often go hand-in-hand. The second pair of thiskind is the support of Open Source software (e.g. Linux) in a company’s applicationsand the development of the Open Source software itself. Especially more complicatedapplications require additional Open Source code to be written until an applicationwill be compatible with the Open Source software.

BMC Software Occasional OS activities.Cooperation with The Open Group to develop an open source Manage-ment Service Broker.URL: http://www.opengroup.org/

EMC No OS activities visible, but development of OS part of job descriptions for currently open positions.

Cadence Design Systems

Major OS activities.TestBuilder C++ testbench class library to be available through Open Source license. URL: http://www.testbuilder.net Engagement in OpenAccess coalition for standard electronic design data-base. URL: http://OpenEDA.org

Adobe Occasional OS activities.Mostly Python plug-ins for Adobe products.URL: http://opensource.adobe.com/

Compamy Open Source activities

Source: The Top 25 Companies are from Software Magazine‘s 2001 Software 500. Their open source activity has been researched by Berlecon Research in March 2002.



12

IBM

IBM’s bet on a standardisedLinux replacing many

Unixes

IBM is active in a variety of Open Source projects. Within the Linux TechnologyCenter alone, around 70 Linux-related projects are listed toward which IBM contrib-utes. This reflects IBM’s strong commitment to Linux: In December 2000, thenCEO Louis Gerstner announced that IBM was planning to spend $1 billion on Linux(Wilcox, 2000). The main motivation behind this was the belief that in the mediumterm proprietary Unix systems would loose their market dominance to a unifiedstandardised Linux.

4

IBM’s Linux investmentalready recouped

While the real investment in Linux might have been somewhat lower than stated, ac-cording to industry observers IBM has taken more advantage of Linux and the open-source movement than any of their competitors (Shankland, 2002a). IBM claims tohave almost recouped its investment in the first year through increased sales of soft-ware and systems. Meanwhile IBM has also ported Linux to their mainframe systems.And as a consequence almost all of the mainframe processing capacity sold by IBMin the fourth quarter of 2001 was for Linux (Shankland, 2002b).

Other Open Source projectsconcerned with Java, XML or

Web Services

But IBM is also involved in a few dozen other Open source projects. Many of theseare focused on new technology trends like Java, XML or Web Services. As IBM is sell-ing a large variety of software and services based on these or adjacent emerging stand-ards and technologies (e.g. its application servers and related products from theWebSphere family), its Open Source activity can help to increase mind share for thestandards and thus ultimately for IBM’s products.

Hewlett-Packard

HP active in the field ofLinux and several other

areas

HP’s major activities in Open Source projects are also in the field of Linux. HP is asponsor of the Linux Standards Base (LSB). The LSB's goal is to develop and promotea set of standards that will increase compatibility among Linux distributions and en-able software applications to run on any compliant Linux system. HP is also a mem-ber of the Open Source Development Lab (OSDL, see Computer Associates) and amember of the GNOME foundation (see Sun). HP has in addition provided someenhancements to basic Open Source infrastructure software like Apache, Squid andSamba.

Hewlett-Packard’s OpenSource strategy centres on

Linux

Hewlett-Packard’s Open Source strategy centres on Linux (Shacklett, 2001). HPconsiders Linux to be the optimal choice of operating system under certain circum-stances. Much of HP’s sale is for solutions, i.e., the combination of hardware, oper-ating system, possibly additional software, and services. Within these packages, theoperating system is only one element. As HP is in addition providing its customerswith solutions across several platforms, HP also has to provide Linux for those marketsegments where it is demanded.

5

This involves the development of software compo-nents that optimise the use of Linux on HP’s hardware.

HP learned early tocombine their products

with operating systems itdoes not own

IDC (2001) points out that HP has an advantage compared to other hardware man-ufacturers for incorporating Linux into their solutions. According to them, HP haslearned early to sell their hardware and solutions together with operating systems itdoes not own (e.g. with Microsoft Windows) – quite contrary to other companies like

4. So far, Linux itself is not fully standardised. There are different distributors (RedHat, SuSE,Caldera, etc.), which combine different pieces of software into their Linux distribution. Alt-hough the Linux kernel is identical, there are many differences in the remainder. This gave riseto several Linux standardisation projects like the Linux Standards Base or Unified Linux, thelatter being the latest initiative.

5. HP: HP’s Linux Strategy, http://www.hp.com/products1/linux/linux_strategy.html.

13



Sun, who have until recently shipped their hardware only with their own operatingsystem.

Compaq

Compaq’s OS projects provide Linux components for its hardware

Compaq is hosting a variety of Open Source projects for software that runs on Com-paq computers. Of the 19 projects listed, 14 are related to Linux software and oftenprovide special drivers or utilities for Compaq hardware. Some non-Linux projectsare concerned with clustering, others provide general utilities. Most projects directlyprofit Compaq which, according to it’s own account, has a very strong position inproviding hardware for Linux-based servers (Compaq, 2001).

SAP

SAP made database available under GNU public license

SAP has made its database, SAP DB, available under the GNU public license. SAPDB is an open, SQL-based, relational database system that provides high availabilityand performance scaling from small to very large implementations. This move can beseen as a strategy to increase the databases market share among SAP users (Hurd,2001). Most SAP using companies do not use SAP DB but rather products from Or-acle, IBM or Microsoft. With SAP DB being Open Source, they have additional in-centives to use SAP DB, e.g. due to no upfront costs (they have to pay a support fee,though), an increased pool of trained programmers, better security due to opensource code and an increasing number of third party support tools.

Support of Linux in expectation of unified operating system

SAP also hosts several programmers from different companies in its so-called SAPLinux lab with the purpose to ensure that its platform mySAP is running on Linux.The lab organises the release of mySAP on Linux and processes Linux-specific sup-port problems. The lab works closely together with the Linux community in optimis-ing Linux. According to SAP, its support for Linux has several reasons: Many SAPcustomers want to combine Intel hardware and their Unix knowledge. At the sametime porting SAP to Linux was relatively easy. In addition, SAP currently has to sup-port a variety of Unix flavours. For this purpose, SAP has to maintain, test and sup-port each possible combination of OS, database, hardware and mySAP release. Aconsolidation of these variants would in their opinion be beneficial to consumers andproducers alike. SAP regards Linux as having a realistic chance of becoming a unifiedUnix running on different platforms.

6

Computer Associates

ca sponsors Open Source Development Lab

Computer Associates (ca) is one of the sponsors of the Open Source DevelopmentLab (OSDL),

7

an initiative to provide Open Source developers with computing re-sources to build data centre and telco class enhancements into Linux and its OpenSource software stack, enabling Linux to become the leading Unix Operating Systemfor e-business development and deployment.

Product sales would benefit from strong Linux as envisioned by OSDL

Computer Associates markets several products running on Linux. Most of these areenterprise solutions, often for enterprise infrastructure management or enterprise in-formation management. The Linux user group targeted with the OSDL initiativewould also be the user group for these products. Therefore ca has an interest in wid-

6. SAP: mySAP Technology on Linux, Frequently Asked Questions, http://www.sap.com/soluti-ons/technology/linux/faq/tech_faq.asp.

7. The other sponsors are a group of Linux and hardware companies: Alcatel, Caldera, Cisco,Covalent, Dell, Fujitsu, Hitachi, Intel, Linuxcare, Miracle, Mitsubishi, Montavista, NEC,Nokia, SuSE, Toshiba, Turbolinux and VA Software.



14

ening the number of Linux instalments for which they can deliver additional soft-ware.

Hitachi

Hitachi supportsinternationalisation of

Linux

Hitachi is mostly involved in Linux development. Six of the seven projects listed onthe web site are concerned with Linux. Some of them focus on Linux for Hitachihardware, others provide tools for Linux development or help in internationalisingthe operating system. (Hitachi is Japanese, and support for Japanese Kanji charactersin software has for a long time been a difficult issue.) Hitachi is also a sponsor of theOpen Source Development Lab (OSDL).

Sun Microsystems

Sun actively contributes toa variety of free and open

source projects

Sun actively contributes to a variety of free and open source projects, including:OpenOffice (the Open Source version of its office suite StarOffice), GNOME,Mozilla, Apache, NetBeans, X Window, WBEMsource Initiative, the University ofMichigan NFS version 4 Linux port, the Grid Engine Project, and Project JXTA. Inaddition, Sun has recently raised its bets on the Linux operating system and will shipan increasing variety of hardware with Linux (Sun, 2002).

Diverse set of motivationsbehind Sun’s OS activities

These projects are rather diverse and so are the motivations behind Sun’s engagementin Open Source projects. Buying the proprietary StarOffice and later releasing it asOpen Source software OpenOffice, e.g., reportedly was mainly a marketing move toincrease Sun’s reputation in the Open Source community and at the same time weak-ening Microsoft (McMillan, 2001).

Some Open Sourceinitiatives support parts of

Sun’s operating system

Other Open Source initiatives centre on products that are part of Sun’s operating sys-tem. X Window is an example for this, where Sun is active in X.org, the consortiumempowered with the stewardship and collaborative development of the X Windowsystem technology and standards. Sun’s engagement in the development ofGNOME, a desktop for Linux, which will replace Sun’s Common Desktop Environ-ment (CDE) on its Solaris-equipped computers, is another example. These graphicaluser interfaces are relatively complex, difficult to develop and maintain, but have alow value within the total product package “computer + operating system” – at leastin the server market which Sun mainly targets. Therefore it is reasonable for Sun notto devote own resources to its further development but rather use a widespread OpenSource product. If fostering its development at the same time helps in making Linuxa better desktop alternative to Microsoft Windows, this weakening of Microsoft’s po-sition would only be in Sun’s interest.

Sun’s Java activities

A third group of Open Source projects evolve around Java, the programming lan-guage controlled by Sun. As Java is increasingly adopted by programmers and becom-ing the major language for writing Internet-based enterprise applications, Sun profitsfrom own Java-based products as well as from license revenue for certain Java usageforms. Therefore it is in the interest of Sun to foster acceptance of Java and to extendits usage into new regions. Open Source projects helping to achieve this are, e.g., Net-Beans, an Open Source platform and Integrated Development Environment (IDE)for Java programs or the Tomcat reference implementation for Java Server Pages, nowpart of the Apache project.

15



Cadence Design Systems

Cadence sponsors industry-specific Open Source projects

Cadence Design Systems is producing tools for the design of chips. The companysponsors two open source projects. It provides Testbuilder, a class library, which en-ables C/C++ to be used as an effective testbench language, through an Open Sourcelicense. A special web site (www.testbuilder.net) has been set up to host software anddiscussions. Cadence also sponsors OpenEDA (openeda.org), an OpenAccess coali-tion for a standard electronic design database. The reason for this sponsorship arehoped-for productivity increases as, according to the consortium, design productivityis a crucial, limiting factor in creating integrated circuits. The goals of the consortiumare to “provide an open standard for IC design data access, along with the supportingsoftware and to gain adoption of the standard within the EDA industry and universityresearch community.”

8

This project can be seen as a classical standardisation effort,which is accompanied by software. As major players of the industry are involved inthis project, a code of conduct has been set up to comply with antirust laws.

8. OpenEDA: OpenAccess Overview: http://www.openeda.org/openaccess_overview.html.



16

17


3 Motivations for firms‘ Open Source activities


Most activity is for development of Linux

Summarizing the results of section 2 leads to the observation that most OS activitiesby the major software makers are in the field of Linux. Even those companies that donot visibly contribute to OS development are in many cases at least passive Linux sup-porters by having ported some of their software to the Open Source operating system.Compared to Linux, other Open Source projects are of much lesser importance. Atleast this is the case for those large software companies investigated; the behaviour ofsmaller, more specialised software companies might differ.

Justification for OSS engagement must come from complementary goods or indirect effects

As firms typically have the target to make profits and as they cannot earn income di-rectly from selling the Open Source software they produce, the justification for theOSS engagements must come in some way from complementary goods or other in-direct effects. The economics literature points out especially the strategy to sell com-plementary products. RedHat, SuSE and the other Linux distributors are goodexamples for companies providing additional products and services related to Linux.Selling additional hardware, as IBM does, is another example. Lerner and Tirole(2002) call such a strategy “reactive”.

Proactively releasing software as Open Source

As they point out, it also makes sense under certain circumstance for a company toreact more “proactively”, e.g. by releasing its software as Open Source. This is the caseif the company can expect to boost profits in a complimentary segment by doing so.As additional condition the profit increase in the complimentary segment must belarger than the profit that could have been made in the primary segment had the soft-ware not been converted to Open Source.

Not all Open Source strategies of software companies successful

This sounds like a fairly straightforward strategy, and as a consequence several com-panies started Open Source projects during the last years. However, as the focus ofthe large software companies‘ activity on Linux shows, most Open Source projectsdid not get the same attention as Linux did. Most notorious example is Netscape’sdecision of 1998 to make a portion of its browser source code freely available to catchup with Microsoft’s Internet Explorer. Initially, the Mozilla project did evolve veryslowly with a limited number of outside developers taking part. It took until 2002 tomake Mozilla 1.0 available.

9

Therefore not all OSS projects that follow the simple“complementary goods logic” seem to work.

Four motivations behind Open Source activities

Therefore one has to look deeper into the OSS projects as well as into the companies‘motivations to participate, to gain a better understanding about the context in whicha commercial engagement in OSS projects can be justified as well as situations whereit is not the case. Doing so, we have identified four major motivations behind thecompanies‘ Open Source activities. Some of these correspond to the complimentarygoods argument, others go a bit further.

9. The history of the project can be seen at http://www.mozilla.org.



18

These motivations are:

❑

Standardisation: overcoming the ghost of Unix wars,

❑

Open Source software as low-cost component,

❑

Strategic considerations, and

❑

Enabling compatibility.

We go into more detail for these motivations in the following sections.

Standardisation: overcoming the ghost of Unix wars

Several different variants ofUnix exist by now

One of the main reasons for Linux activities of many companies – be it actively con-tributing or passively adopting their software – is the hope to “overcome the ghost ofUnix wars.” Since Unix has forked in an early stage of its development, and sincehardware manufacturers have developed Unix variants specially targeted to theirhardware, several different variants of Unix exist by now. Examples are Sun’s Solaris,IBM’s AIX, HP’s HP-UX, Silicon Graphics’ Irix, etc.

Keeping basic operatingsystem up-to-date is costly

With the key selling arguments for hardware moving away from the operating systemtowards higher-level software support (e.g. application server, Java) and towards moregeneral features (e.g., reliability, security, low total cost of ownership), neither do cus-tomers want to worry about the operating system nor do hardware manufacturerswant to spend considerable resources on keeping their basic operating system up-to-date, as this becomes less important for gaining a competitive advantage. Also soft-ware manufacturers have to support an increasing number of combinations betweenoperating systems and hardware and would love to reduce the necessary effort for thistask.

Common interest in favourof a single operating

system

Thus, there is a common interest in favour of a single operating system providing ba-sic functionalities within the package of hardware, software and services that thesecompanies sell. For every party involved this would be advantageous. It would de-crease the cost of operating system maintenance by sharing the cost for introducingadditional functionality. It would decrease the support costs for other software man-ufacturers and it would decrease the risk for customers to bet on the wrong operatingsystem, which might become unsupported at a later point in time.

Open Source foundationallows focus on other

elements of product bundle

In addition, companies can focus on other elements in the product bundle they offer,i.e. focus on their core competencies. Especially for large software companies, theseare typically more in creating solutions to actual business or IT problems and in mar-keting these than in providing basic computing functionality. By focusing on theircore competencies and integrating the basic OS operating system into their productbundles, they can also speed up their rate of innovations.

GPL license ensures thatfoundation cannot be

hijacked…

An Open Source operating system is well suited for this purpose as its license modelensures open access to the software as well as the possibility for everyone to influenceand participate in its development. Due to its viral GPL license Linux is well suitedfor such a role as basic foundation as it cannot be “hijacked” by any party, at least notback into a proprietary software. Any derived works must be made available again un-der the GPL license. Problem of the GPL is, however, that it strictly limits integrationinto other software.

… but influence dependson resources devoted to

project

Nevertheless, in practice the influence about the project’s future development is big-ger, the more resources can be devoted to participating in the project. Therefore it isespecially interesting for large companies like IBM to become active participants inthe Open Source development process to shape it in their interest. This is the samebehaviour as can be observed by standard-setting organizations. OS development

19



processes are open for participation by everybody, but they are not democratic in aone-man-one-vote sense.

Open Source software as low-cost component

Product bundles containing Open Source cheaper than fully proprietary bundles

A second reason for the increasing usage of Open Source software, especially Linux,as an input into the production of product bundles and, in some companies, increas-ing participation in its development, is simply its lower cost. With the basic operatingsystem itself being not any more the key property to distinguish products, the abilityto produce bundles or hardware and software meeting specific requirements at lowcost is a competitive advantage. The combination of relatively cheap Intel hardwareand a free operating system fits these requirements better than an expensive albeitpowerful proprietary hardware combined with a proprietary operating system.

Development is to a large extent customer driven

This recent development is to a large extent customer driven. Customers have expe-rienced that the rapidly increasing IT-requirements for processing e-business transac-tions can not only be met by buying highly powerful expensive solutions (verticalscaling) but also by combining a high number of cheap Intel/Linux-systems (horizon-tal scaling). But if they want to combine several dozen machines to a cluster, the li-cense fee for a traditional operating system becomes a significant cost factor. This canbe avoided by using Linux. The same argument applies to another range of recentlyintroduced products, the so-called appliances. These are combinations of hardwareand software for a single purpose (e.g., firewall, email-server), which are often posi-tioned in the low-cost segment.

Strategic considerations

Release of software as Open Source in non-essential areas

While most OS activities are in the Linux field, some companies also contribute to avariety of other OS projects or have even released their software to the Open Sourcecommunity, like SAP did with its SAP DB database. Typically, just like the Linux en-gagements, these activities are in areas which provide a rather basic functionality forsolutions built on top of it than in the development of actual products. SAP, e.g., isproducing and selling enterprise resource planning (ERP) software. A database is onlya precondition for running such software. Likewise, Sun is selling complete server so-lutions. The X Window window system or the GNOME desktop is only one elementof this solution and not even an important selling point.

10

Releasing software as Open Source can weaken competitor

But for other companies, which are in some areas competitors to SAP or Sun, theseOpen Source projects are direct competition. Oracle, for example, sells databases aswell as ERP-systems. By releasing the SAP DB as open source – which makes it muchcheaper for users – SAP can hope to snatch market share from Oracle in the databasefield and thereby weakening Oracle as a whole, possibly making it a less strong com-petitor in the ERP field, too. The same applies to Sun. By helping the improvementof GNOME, Sun raises the chances that Linux (which can also use GNOME) be-comes a serious operating system for the desktop. Sun itself is not really active in thedesktop computer business, but its archrival and competitor in the server business,Microsoft, is. This strategy is even more pronounced in Sun’s support of theOpenOffice project, which is also a potential threat to Microsoft.

10. An exception is OpenOffice. However, as noted above, the move to make this software availableas Open Source can be seen as strategy to please the Open Source community.



20

Enabling compatibility

OS projects to make ownsoftware or hardware

compatible with OSsoftware

Some companies show only limited OS activities, mainly by making their software orhardware compatible with OS software. For example, Adobe has released some Py-thon plug-ins for its products. Another example are several hardware makers like IBMor HP, which have released OS software that enables Linux to run on their computersand to make optimal use of special hardware features built into their products.

Viral licenses may force thecompanies to provide Open

Source software

As long as the respective OS software is licensed under a viral OS license (e.g. GPL)and the company‘s OS-related software incorporates parts of the OS software (e.g. li-braries under certain circumstances), the company has to provide its OS-related soft-ware as OS again. But even if this is not the case, marketing considerations might besufficiently strong reasons for releasing the source code. Obviously, this makes onlysense when the OS code makes only use of publicly known APIs of the proprietaryhard- or software. If the company does not want to make these known, it would notlet itself being forced to do so by an Open Source license.

Motivations of Open Source opponents

Most of the non-contributors are probably

free-riders

Only about half of the companies investigated engage in Open Source developmentprojects. It is therefore useful to ask why others do not contribute. Most of the non-contributors are probably simply free-riders, who use the OS software – especiallyLinux – but do not contribute to its further development. While this may be morallyobjectionable it is a perfectly rational behaviour if the company can make direct useof the open source software as it is.

Objection to OS model forprincipal reasons

There are other companies, though, that object the OS development model for prin-cipal reasons. Microsoft is probably the most prominent one, which has proposed analternative licensing model called “shared source” where some companies and otherinstitutions like universities get so see selected parts of the source code. While thereare several high-brow principal arguments provided by Microsoft and others againstthe Open Source model, there are also plain business reasons for some companies toobject to the Open Source model.

Open Source componentsincompatible with strategy

of tight integration

The most pressing business argument is that some OS software is or can become animmediate competition to products offered by these companies. Linux in combina-tion with other OS software is a competition to Microsoft’s products, as the debateabout Linux in the German Bundestag has shown vividly. Also, Microsoft has decid-ed to integrate all its software components as well as coming services like .NET verytightly. Within this strategy, there is no place for certain components – for examplethe server operating system – to be something other than a Microsoft product (Wil-cox and Shankland, 2001).

Microsoft not the only OSopponent

But Microsoft is not the only company opposed to OS software when it comes tocompeting products. JBoss, for example, an OS Java application server, is a directcompetition to some of BEA’s products. Therefore BEA made clear that according totheir opinion Open Source is not suited to mission-critical systems but rather a vehi-cle for low-end non-critical systems (Coleman, 2001). This is an assessment that thesponsors of the Open Source Development Lab, which explicitly has the goal to makeLinux an operating system for large critical systems, certainly would not share.

Objection to Open Sourcefor fear of hijacking

Finally, some companies object to make their software available as Open Source soft-ware, since they fear that its further development process could be hijacked by others.Sun’s hesitance of giving third parties more saying into the future development ofJava belongs into this category, as there exists the potential threat of Microsoft engag-ing too much in this activity. This example also shows that companies can take dif-ferent positions for different products, depending on the actual circumstances.

21



Conclusions

Companies take different positions towards Open Source

From these different motivations for and against engagement in Open Source activi-ties it becomes clear that companies take different positions towards Open Source.These positions depend on the kind of software in question and on its relation to thecompanies‘ major products. As soon as these are in direct competition, the compa-nies‘ position towards Open Source is much less favourable than in a setting, wherethey are complements or where the (cost-saving) Open Source software replaces a partof the product package that is not considered core business.

Short-run and long-run motivations

The motivations for engaging in Open Source projects can be considered to be of twokinds: One set of motivations is rather short-run and focused on single software prod-ucts. Making certain software packages available as Open Source for strategic reasonsor developing Open Source programs to make one’s own software compatible withOpen Source software belong into this category. The second set of motivations islong-run in comparison and focused on Open Source components within largerpackages of hardware and software. Helping to let Linux become a unified Unix op-erating system and replacing own software components with Open Source ones be-long into this category.

Open Source software as basic infrastructure

It is the latter motivation that is often behind the analysed companies‘ engagement inLinux development. Such a long-term goal of using Open Source components as ba-sic building blocks – very much in the sense of infrastructure upon which those prod-ucts are build that actually compete in the market – has several analogies in other areasof business. These analogies are explored in the following section.



22

23


4 Policy implications of Open Source activities

4 Policy implications of Open Source activities

Standardisation and basic research activities as analogies

The economics literature suggests two areas that resemble in some way the engage-ment of firms in development of infrastructure-type Open Source software. The rel-evant areas are the economic analysis of standardisation as well as the economicanalysis of basic research by companies. The parallels between these two areas and theanalysis of Open Source activity will be drawn in this section.

Framework to discuss policy issues

These two frameworks can also be used to analyse the relationship between OpenSource activity, regulation and public policy. This is done on a rather principal leveland focused on indirect government policies. These are activities by government bod-ies and regulatory authorities aimed at setting the right business or legal frameworkwithin their entities. They include competition policy and the protection of intellec-tual property rights. The analysis of direct policy measures to foster use and develop-ment of Open Source has already been conducted extensively by others (e.g., Schmidtand Schnitzer, 2002; Evans and Reddy, 2002).

4.1 Open Source activity as standardisation effort

Aim of unified Unix resembles standardisation

As pointed out in the previous section, the main Open Source activities of firms arein the field of Linux. A major motivation for fostering its further development is thepotential reduction of Unix variants towards one major Unix operating system andthe associated cost savings. This is actually a motivation that is shared by many otherstandardisation efforts.11 Standards provide better interoperability (a program forLinux on IBM can – more or less – run on Linux for HP) and thus a larger marketfor additional products. They also lead to more trained personnel being available forthis operating system and thus lower costs for this personnel.

Standardisation reduces investment risk

In addition, reduction of variety as one outcome of standardisation reduces also therisk of investments. If Linux becomes the major Unix OS, developing an applicationfor Linux is in the long-run a safer bet than developing it for, let’s say, Sun’s Solaris.This point is especially important for small companies, as betting on the wrong op-erating system can force them to close shop. The opportunities of a widely acceptedoperating system for small companies are shown by the considerable number of SMEsthat develop applications for Microsoft Windows, a standard that is also expected tostay.

Standardisation can make the future path of product innovation more focused

Finally, standardisation can make the future path of product innovation more fo-cused. While innovation in not-standardised environments often resembles more abush – for each branch, there are several innovations leading to new branches – thesequence of innovations in standardised environments resembles more closely a well-

11. For an overview of the economic aspects of standardisation see, for example, Swann (2000).

4.1 Policy implications of Open Source activities


24

trimmed tree with a strong trunk. This requires, however, that the standard evolvesover time and incorporates the best of competing innovations into the standard set.

Motivations forparticipating in

standardisation process

As in all other forms of standardisation, it is rational for companies to participate inthe standardisation process mainly for two reasons. First or all, if the proposed stand-ard indeed becomes an accepted standard, they have a head-start before other com-panies that did not participate to the same extent. And secondly, they can try toinfluence the standardisation process in their particular interest, for example bybringing the standard as closely as possible to their existing technology. It is thisstandardisation process and the participation issues that raise the most questions forpublic policy and regulation. Two items are particularly important, competition pol-icy and issues of intellectual property rights.

Implications for competition policy

Standardisation efforts atwo-edged sword

For competition policy standardisation efforts are a two-edged sword. On the onehand, standardisation decreases the barriers for entry into a market and levels theplaying field, which is favourable to competition. Such group standards might inten-sify the competition in other product markets that use the standard as basis. Exampleslike appliances or other packaged products using Linux point into this direction. Onthe other hand standardisation bodies and their activities do provide a potential forcollusive behaviour among those participating, which can decrease competition.Therefore competition authorities watch standardisation efforts very closely.

Openness for participationtypically the case in Open

Source projects

There are two major points of importance here. The first is openness for participa-tion. If everybody can join the club on non-discriminatory terms, the probability thatthe standardisation process is used to keep others out of the market is lower. On theother hand, if access to the standards body is restricted, the activities might well beworth further investigation. Open Source activities do not pose much of a threat re-garding this issue. As everybody can join Open Source activities and as also the out-come of the activities is available to everyone, it is very unlikely that they pose a threatto competition.

Abuse of OS projects toexchange sensitive

information also unlikely

The second issue is the potential abuse of a standardisation body as a platform for ex-changing sensitive information with the aim to reduce competition. Pricing agree-ments are the most obvious kind of forbidden behaviour. Exchanging sensitiveinformation regarding targeted markets, planned strategies, etc. is similarly problem-atic. Again Open Source projects pose less of a threat to competition than otherstandardisation efforts. Open Source activities typically take place in an open space,i.e., on the Internet, and not in smoke-filled backrooms.

As they are open to everyone and as most of the information flow in form of e-mailsand group discussions takes place in written form, is archived and visible to most ofthe world, the incentives to abuse the Open Source projects for such anticompetitivebehaviour are rather low. By the same token, absence of such an open communicationcould be a signal for authorities to watch the activities more closely.

Code of conduct withinOpen Source projects

To clarify this issue and define appropriate behaviour some initiatives, e.g. OpenE-DA, have given themselves a code of conduct explicitly forbidding the misuse of theinitiative for anti-competitive activities.

Competition issues ofrestricted shared source

initiatives to beinvestigated

Somewhat more problematic from the point-of-view of competition policy might beshared source initiatives like Microsoft’s, where access to the source code is limited tocertain kinds of companies or institutions. This becomes especially relevant if thecode is considered to be of essential importance, e.g. because it defines an operatingsystem with large market share. If not everyone has access to the provided informa-

25


4.1 Open Source activity as standardisation effort

tion or only under rather strong restrictions, some might judge such activities as beingdetrimental to competition.12 However, up to now there has not been much discus-sion of these issues. Therefore it is too early for a final assessment.

Issues related to intellectual property rights

Hijacking standardisation group problematic issue

A further issue arising in classical standardisation activities is the possibility that a par-ticipant tries to hijack a standardisation group. Typically this works in the way thatthis party participates in a standardisation body, agrees to a standard, helps in pro-moting it and eventually, when the standard becomes widely accepted, pulls a patentcovering part of the standard out of his pocket. There do exist a few examples for suchbehaviour, like Unisys’ discovery of owning the rights for the LZW compression al-gorithm employed in the GIF graphic format or Dell’s patent covering part of theVesa Local Bus standard (Lemley, 1996). It is not completely clear, whether the be-haviour of these companies was intentional from the beginning or whether the com-panies only by accident noticed that they are sitting on a potential IPR goldmine.While these examples concern hardware and algorithms, the increasing importanceof software patents also makes this problem more acute for Open Source projects likeLinux.

Assurance of participants that no intellectual property claims will be made

To avoid such potential problems, there is the possibility of imposing that participat-ing in the standardisation process requires an assurance of every participant that hedoes not own any intellectual property right related to this standard. The Internet En-gineering Task Force IETF has chosen this approach. Others like ANSI require onlythat access to the intellectual property is granted on fair conditions (Lemley, 1996).Recently this issue has led to a vivid discussion within the World Wide Web Consor-tium W3C.13

Open Source projects pose several problems to intellectual property clearance

Such a rule can be followed comparatively easily in small groups of participants. InOpen Source projects, where often many participants are involved, who in additionmight move into and out of the projects running for several years and in the processmight even change their employers, it becomes rather difficult to handle. In additionthe participants in Open Source projects are typically people, even if they are paid bytheir employers to contribute, while the owners of patents are mostly companies. Itis unclear, under what circumstances the participating individuals can provide legallybinding statements in the form of those given above.

Problem is likely to get more attention in the future

With the increasing acceptance of software patents also in the European Union, thispotential problem is likely to get more attention – also for Open Source projects. Theissue has been addressed to some extent in OS licenses, for example in the GPL. Sec-tion 7 states that if due to patent infringement claims conditions are imposed on dis-tributing a software that contradict the terms of the GPL (e.g., royalty payments), thesoftware must not be distributed.14 As the GPL is viral, this obviously leads to prob-lems if patent claims are made for very basic pieces of software, which are included inseveral other derived works. The consequences of this problem – in theory as well asin reality – are still to be investigated.

12. For example, Microsoft’s OEM (Original Equipment Manufacturers) Source Licensing Pro-gram is only open to such OEMs that “[m]eet the OEM group's definition of ‘Multinational’or ‘Datacenter OEM,’ or be a Windows CPU vendor.” Cf. http://www.microsoft.com/licensing/sharedsource/oem.asp.

13. http://www.w3c.org/2001/ppwg/.14. http://www.gnu.org/copyleft/gpl.html.



26

The role of government for Open Source as standardisation

Do typicalrecommendations aboutstandardisation apply to

Open Source?

As standardisation is an activity that has taken place for quite some time in severaldifferent areas, there are several opinions about what the government should orshould not do with respect to such standardisation activities. It is worth asking,whether the typical recommendations also apply to Open Source activities when in-terpreted as standardisation.

Is level of standardisationoptimal?

The first, and probably most debated issue is, whether the amount of standardisationactivity taking place is optimal, given that standardisation has positive implicationsfor competition and thus for consumer welfare. Economic theory that could providesome insight, however, has been more concerned with standards races, where several(at least partially) incompatible standards compete for market dominance and hastried to answer questions about fair and unfair practices in these races and about thedeterminants of the optimal outcome.

Government should becareful when pushing

companies intostandardisation-focused

Open Source activities

An assessment about the optimal degree of standardisation is rather difficult, as con-sumers’ utility from variety, the importance of variety as driver of technical progressand the static efficiency gains from the use of standards have to be balanced somehow.The conclusion for government policy would be that it is hard to assess the optimaldegree of standardisation. There are no convincing arguments, why governmentshould be able to determine this optimal level better than the market does. Thereforeit should be very careful when pushing companies into standardisation-focused OpenSource activities – as tempting as the vision of a fully standardised Open Source tech-nology core and the associated static gains may be.

Government should notpick standards

What frequently happens in traditional standardisation processes is that the processstalls and participants cannot agree to a standard. In such situations it is sometimessaid that government should choose a standard, as any standard is said to be betterthan no standard. Mobile communication in Europe, which is based on the singleGSM standard, is often cited as an example for the benefits of standardisation. How-ever, a closer look at this example shows that it also had its drawbacks: Japan, for ex-ample, has a much more advanced mobile phone system and a much larger variety ofmobile data services than Europe, despite several incompatible standards competingwith each other. Therefore public policy would have to be very careful in taking sidesand trying to push development in one or another direction.

Balance participation andrepresent excluded

interests

A further claim about government’s role in the standardisation process is that itshould try to balance participation and represent excluded interests. This claim is alsovery much influenced by traditional standardisation procedures where industry-heav-yweights agree about the future course of technology. With respect to Open Sourceprojects like Linux, though, participation is in principle open to everybody. Thus, therole of government is much smaller here than in traditional standardisation efforts.

Are European interestsunderrepresented in Open

Source projects?

One potentially underrepresented group might be Europeans. It has been arguedabove that especially large software companies have an incentive as well as sufficientresources for participating in or even for driving Open Source projects. As these com-panies are mainly located in the United States and as the US market is larger than thatof any single European country, it might be the case that US conventions – language,character sets, business rules – are better represented in the resulting software thanEuropean ones. This problem is unlikely to be very large, though. As the FLOSS de-veloper survey showed, European programmers make up the large majority of thosedevelopers surveyed (Ghosh et al., 2002). Even taking into account that the survey isprobably biased towards European developers, there remains a substantial number ofEuropean Open Source developers.

27


4.2 Open Source activity as basic research

Education role remainsTaking all these points together from the point of view of Open Source projects asstandardisation activities, the most important role of government is probably to edu-cate companies about this role OS projects can have and to make clear what thechances of participation and the potential threats of ignoring these activities are. Itmight also be useful to investigate whether other Open Source activities besides Linuxare standards-like and point out these opportunities to companies without forcingthem to participate.


Open Source engagement similar to basic research

Besides the interpretation of OS activity as standardisation activity, there is a secondreasonable analogy that can be drawn. The engagement of software manufacturers inOpen Source projects is very similar to firms‘ engagement in basic research. In bothcases the immediate gain from this activity seems to be much smaller than from de-voting the resources to the development of other products that can be sold directly.

Basic research by firms with positive spillover effects

The economics literature has investigated thoroughly why firms nevertheless invest inbasic research (e.g., Rosenberg, 1990). This question is of huge importance for policymakers, since the social returns from basic research, for example in form of growtheffects, are considered to be extraordinarily large (Griliches, 1996). The same is oftensaid about Open Source software. As everybody can learn from reading this sourcecode and using it as a basis for own software, positive externalities of these OpenSource activities exist.

Engagement in basic research when some returns can be captured

The major arguments from the economics literature explaining basic research activi-ties have already been met during the analysis of the OS projects from the largest soft-ware companies. For example, it is said that an engagement in basic research isrational, when firms can capture at least some of the gains from this activity. They donot need to capture the entire returns from basic research. Indeed, this would by so-cially inferior, as it is especially the externalities that make basic research desirable.There are several ways in which a company could capture some of the benefits: Forexample, it can use it in combination with other inputs for new products (just likeLinux on IBM mainframes) or it can try to obtain first mover advantages. These ar-guments certainly apply for Open Source projects.

Knowledge from basic research can best be understood when companies are involved in research process

A further reason for basic research activities is that knowledge from basic research canonly be understood and thus be incorporated in innovations, when companies are fa-miliar with the research process, the jargon used, etc. This knowledge is best gainedby employing at least a few people with strong ties to the scientific community.Knowledge is not, as it is sometimes argues, “lying on the shelf”. The same might betrue to some extent for Open Source projects. While much Open Source software caneasily be used as is without being involved in its production process, incorporatingOpen Source software in one’s products requires a deeper understanding of its capa-bilities and problems, which is best gained by participating in its development proc-ess.

Issues related to intellectual property rights

IPR protection raises incentives for basic research

Economists have long since argued that society would invest insufficient resources inbasic research in all market regimes. While monopolists do have a lower incentive toinnovate than firms in perfect competition, the latter have the problem that a com-petitor can quickly exploit the useful new knowledge when it is unprotected. The ob-vious solution to this problem is IPR protection, e.g. in the form of patents. However,



28

these introduce another inefficiency. Since the knowledge has been produced and canbe distributed almost without cost, it is inefficient not to do so. This is a dilemmawithout a simple solution.

Observed outcome of OpenSource activity optimal for

firms

As firms are voluntarily choosing to participate in Open Source projects, the observedvolume of Open Source projects by firms can be interpreted as the (local equilibrium)outcome of their research investment. As firms can protect most of their intellectualproperty in the domain of software development, it can be assumed that they onlygive as much intellectual property away in the form of Open Source software as is op-timal for them.

Are Open Sourcepeculiarities keeping

companies from releasingcode as Open Source?

The question to address is thus, whether this level of Open Source activity could beincreased without weakening the rate of innovation within these companies. Arethere certain peculiarities of the Open Source process that keep such firms from par-ticipating that under different regimes would be willing to make their knowledgeavailable to others?

Viral nature of GPL One issue pointed out, for example by Microsoft,15 is the viral nature of the GPL(which governs Linux) and especially ambiguities in its virality, which supposedlymakes it difficult to build commercial software on top of Open Source software.While a discussion of this legal issue is beyond the scope of this paper, it has to betaken in account that unclear legal implications might indeed be issues keeping com-panies from taking part in those Open Source projects governed by such licenses orfrom including such software as infrastructure components into their products.

However, these ambiguities do not concern the release of formerly proprietary soft-ware as Open Source, as companies are free to choose the license they want when do-ing so. The strictness of the GPL is one reason why many software companies useBSD Unix, governed by the more liberal BSD license, as foundation for their com-mercial software.16

The role of government for Open Source as basic research

Basic research often limitedto certain sectors and large

firms

Empirical research about basic research has frequently shown that basic research istypically limited to very few sectors. Within these sectors there exists a handful offirms, typically large firms, that dominate the basic research picture (Rosenberg,1990).

Issue seems lesspronounced in Open Source

However, this issue seems to be less pronounced in the Open Source area. For exam-ple, the hurdles for participation (barriers to entry) in Open Source projects are lowerthan for conducting classical basic research, as only smart people are needed and nocostly research infrastructure. Indeed, the FLOSS user survey does not show signifi-cant differences between large and small companies in letting their developers pursueOpen Source projects on company time (Berlecon Research, 2002). Nevertheless,there might be differences between large and small companies in deliberate decisionsto participate actively in Open Source projects for reasons like those set out above.

Often large companies cancapture indirect returns

better than smaller ones

A major hypothesis trying to explain the dominance of large companies in basic re-search is that their product diversity as well as their sales and marketing power raisesthe potential of being able to use the research outcome. Small companies in compar-ison need to target their research investment much more to their immediate needs.This argument also applies to participation in Open Source activities. As companies

15. See the discussion with Microsoft’s Craig Mundie at the O’Reilly Open Source conference inJuly 2001.http://linux.oreillynet.com/pub/a/linux/2001/08/09/oscon_panel.html

16. Also the Apache project uses a BSD license explicitly for these reasons.

29



cannot earn directly from these activities, they must find a way to capture some of thegains in an indirect way. Large companies can do this often better than small ones,although there do exist counterexamples like Sendmail, a company selling messagingsolutions on the basis of software components it released as Open Source.

If SMEs participate less than government should create participation incentives for them

Thus, the extent to which the Open Source activities of small and large companiesdiffer, has to be investigated in more detail. If SMEs participate less than optimal inthese activities, then for government policy the same arguments apply for supportingOpen Source engagement than for supporting (other) basic research. Policy shouldcreate incentives to increase this participation. Not only would this give smaller com-panies similar benefits as larger companies, but it would also increase the amount ofOpen Source software available and would thus most likely raise the rate of technicalprogress in software manufacturing.

Should government conduct basic research?

With respect to basic research it is sometimes suggested that the government itselfshould become active and conduct basic research, as the incentives for private entitiesare insufficient. Likewise one could argue that the government should engage inOpen Source activities as the existence of Open Source software is socially beneficialbut the incentives are not sufficient to develop (enough) OS software. However, thegovernment does not have a good record of choosing the right projects, and thereforeit is likely that it will develop Open Source software that does not meet the require-ments of the market but rather those of the people deciding about the project.17

Support of Open Source development justified with caveats

A different issue would be the support of Open Source development. To the extentthat private entities do not engage sufficiently in Open Source activities, support ofthe latter is justified provided that three issues are taken into account. First of all, thesupport should be structured in a way that lets the market decide about whichprojects are useful. This is, e.g., the case when supporting infrastructure for OpenSource development. Secondly, support should go to those kinds of Open Sourceprojects that provide software closest to basic research, i.e. infrastructure-like softwarethat can be used as component in many other kinds of software. And thirdly, the li-cense regime of the supported Open Source software projects should be such that theresults can be used in as many ways as possible. This would exclude strong viral licenseregimes such as the GPL.

17. The debate about the use of Linux in public institutions illustrates the problem: Linux is notthe only alternative to Microsoft products (BSD-based products and Macintosh are other alter-natives) and it is not clear which will prevail in the long-run. Initiatives forcing public instituti-ons to use Linux might put them on a sub-optimal technological trajectory in the long-run.



30

31


5 Conclusions

5 Conclusions

Profit-maximising large software companies actively contribute to Open Source projects

A major observation in this paper is that profit-maximising large software companiesdo actively contribute to Open Source projects. Not all of them do so, not all do it tothe same extent, and the motivation for their activity is obviously purely selfish. Nev-ertheless, by doing so they contribute to the amount of Open Source software avail-able to everybody and thereby contribute to economic growth and social welfare.

Motivations bear resemblance to standardisation and basic research

The reasons for doing so bear some resemblance to other activities of companies likethe engagement in standardisation efforts or the conduct of basic research. Thus, sim-ilar issues related to competition policy or intellectual property rights can be shownto play a role in Open Source projects, although mostly to a smaller extent than inother surroundings.

Possibly sub-optimal level of Open Source activity

There are some arguments implying that the amount of activity in Open Sourceprojects by companies is sub-optimal, as the companies do not fully take into accountthe positive externalities from the availability of the Open Source code. This providesa justification for certain ways of government support to Open Source projects.

Support “with the market”: Removing barriers to Open Source activities by commercial entities

Such support should not be provided “against the market” but rather “with the mar-ket”. Going with the market means trying to understand why firms do contribute toOpen Source software, setting the legal and regulatory framework in a way that allowsthem to do so and educate those that are not yet doing so about the benefits they willhave from contributing to Open Source. It would, however, also mean accepting thatfirms are often more willing to support development of some kinds of Open Sourcesoftware (e.g. for infrastructure- or standards-like software) than for others (e.g. spe-cific applications).

Support “against the market”: Open Source software as free alternative

Going against the market means supporting Open Source projects that provide an(free) alternative to commercial programs, where only one or two manufacturers re-main. This is not necessarily beneficial. As software tends towards a natural monop-oly, creating an artificial duopoly might be sub-optimal. With the same resourcesother projects could be supported that increase the amount of Open Source softwarein other layers of software and thereby benefiting a larger group of users in additionto speeding up the rate of technical progress.

Issues for public support of Open Source activities

Public support of Open Source activities requires the following three issues to be tak-en into account. First of all, the support should be structured in a way that lets themarket decide about which projects are useful. This is, e.g., the case when supportinginfrastructure for Open Source development. Secondly, support should go to thosekinds of Open Source projects that provide software closest to basic research, i.e. in-frastructure-like software that can be used as component in many other kinds of soft-ware. And thirdly, the license regime of the supported Open Source software projectsshould be such that the results can be used in as many ways as possible. This wouldexclude strong viral license regimes such as the GPL.

5 Conclusions


32

33


6 References

6 References

Berlecon Research (2002): FLOSS Final report, Part 2: Use of Open Source Softwarein Firms and Public Institutions: Evidence from Germany, Sweden and UK.

Coleman, Murray (2001): BEA Focuses on Businesses Stuck in the Middle, October2001, http://www.bea.com/press/ibd_qa.shtml.

Compaq (2001): Compaq maintains #1 position in Linux server market,20.12.2001, http://www.compaq.ch/de/press/release.htm?NoteID=651.

Evans, David S.; Bernard Reddy (2002): Government Preferences for PromotingOpen-Source Software: A Solution in Search of a Problem, 21.7.2002, mimeo.

Ghosh, Rishab Aiyer; Ruediger Glott; Bernhard Krieger; Gregorio Robles (2002):FLOSS Developer Level Analysis, working paper, mimeo.

Griliches, Zvi (1996): Productivity, R&D, and basic research at the firm level in the1970’s, The American Economic Review, 76 (1996) 1, 141-154.

Hurd, Jim (2001): Open Source for the Enterprise, 26.7.2001, http://www.extreme-tech.com.

IDC (2001): Expanding the Frontiers of Enterprise Computing with Hewlett-Pack-ard Linux Solutions – An IDC White Paper, August 2001.

Lemley, Mark A. (1996): Antitrust and the Internet Standardization problem, Con-necticut Law Review 28 (1996), 1041.

Lerner, Josh; Jean Tirole (2002): Some Simple Economic of Open Source, Journal ofIndustrial Economics, 50, 197-234.

McMillan, Robert (2001): Sharing the Dot in Dot-com, Linux Magazine, May 2001.

Raymond, Eric (2000): Homesteading the Noosphere: An Introductory Contradic-tion, http://tuxedo.org/~esr/writings/homesteading/homesteading/.

Rosenberg, Nathan (1990): Why do firms do basic research with their own money?Research Policy, 19 (1990), 165-174.

Schmitz, Klaus; Monika Schnitzer (2002): Public Subsidies for Open Source? SomeEconomic Policy Issues of the Software Market, 14.6.2002, mimeo.

Shacklett, Mary (2001): HP open source strategy centers on Linux, HP Chronicle,January/February 2001, http://www.serverworldmagazine.com/hpchronicle/2001/0102/opensource.shtml.

Shankland, Stephen (2002a): IBM: Linux investment nearly recouped, CNETnews.com, 29.1.2002, http://news.com.com/2100-1001-825723.html.

Shankland, Stephen (2002b): IBM puts Linux converts on display, CNETnews.com, 31.1.2002, http://news.com.com/2100-1001-826926.html.

6 References


34

Sun (2002): A Closer Look at Linux, 19.3.2002, http://www.sun.com/2002-0319/feature/.

Swann, G.M. Peter (2000): The Economics of Standardisation – Final Report forStandards and Technical Regulations Directorate, Department of Trade andIndustry, December 2000, mimeo.

Wilcox, Joe (2000): IBM to spend $1 billion on Linux in 2001, CNET news.com,12.12.2000, http://news.com.com/2100-1001-249750.html.

Wilcox, Joe; Stephen Shankland (2001): Why Microsoft is wary of open source.CNET News.com, 18.6.2001, http://news.com.com/2100-1001-268520.html.

Berlin, July 2002



and Business Models Basics of Open Source Software Markets

3

BERLECON RESEARCH GmbH


10117 Berlin

Tel.: +49 30 285296-0

Fax: +49 30 285296-29



Acknowledgements:

This work was prepared by Dorit Spiller and Thorsten Wichmann from BerleconResearch. It is part of the final report for the project „FLOSS – Free/Libre OpenSource Software: Survey and Study“, which was financed under the European Com-mission‘s IST programme, key action 4 as accompanying measure (IST-2000-4.1.1).

Disclaimer:





4

V 1.1 - 020905


5

Table of contents

1 Introduction....................................................................................................... 9

2 Software and the Open Source phenomenon.................................................... 112.1 The Open Source phenomenon .................................................................. 11

2.1.1 Definition of Open Source software ...................................................... 112.1.2 Historic development ............................................................................ 122.1.3 Open Source licensing........................................................................... 142.1.4 Main Open Source products and development projects......................... 16

2.2 Software characteristics and the software value chain ................................... 222.2.1 Software characteristics.......................................................................... 222.2.2 The software value chain: overview........................................................ 232.2.3 Programming and production ............................................................... 242.2.4 Marketing and distribution ................................................................... 262.2.5 Services.................................................................................................. 282.2.6 Software product categories ................................................................... 30

3 Open Source and the software market: market segmentation and hypotheses ... 313.1 The software market.................................................................................... 313.2 Business dynamics and influence of Open Source software .......................... 32

3.2.1 The software products market ............................................................... 323.2.2 The market for software-related services ................................................ 373.2.3 The market for embedded software products and services...................... 38

4 Business models based on Open Source software .............................................. 414.1 Overview..................................................................................................... 414.2 Distributors and retailers ............................................................................. 42

4.2.1 Original Linux distributors.................................................................... 424.2.2 Niche and specialty OSS distributors .................................................... 444.2.3 Retailers of OSS distributions and complementary products ................. 45

4.3 OSS-related services..................................................................................... 464.3.1 OSS development and community enablers .......................................... 464.3.2 OSS-related services and support........................................................... 48

5 References ........................................................................................................ 515.1 Literature..................................................................................................... 515.2 Web resources ............................................................................................. 54

5.2.1 Publications on OSS and Linux............................................................. 545.2.2 Publications on IT in general ................................................................ 555.2.3 Software aggregation ............................................................................. 555.2.4 Organisations/OSS community............................................................. 555.2.5 Commercial OSS companies ................................................................. 565.2.6 OSS projects and products .................................................................... 56


6


7

Figures and tables

Figure 2–1 Classification of OSS versus other software.......................................................... 11Table 2–1 Licenses overview................................................................................................. 16Figure 2–2 Matrix of products and services ........................................................................... 23Figure 2–3 Software value chain ............................................................................................ 23Figure 2–4 Software value chain for standard software versus individual software.................. 24Figure 2–5 Technical and functional software product categories .......................................... 30Figure 3–1 Structure of the IT market................................................................................... 31Figure 3–2 Structure of the software products market ........................................................... 33Figure 3–3 Software for the server and for the desktop/client market .................................... 33Figure 4–1 OSS business models ........................................................................................... 41Figure 4–2 Primary focus of distributors in the software value chain ..................................... 44Figure 4–3 Primary focus of service companies in the software value chain............................ 48Figure 4–4 Areas of success in the OSS-related service business ............................................. 50


8


9

1 Introduction

1 Introduction

Background contribution to FLOSS project

This third part of the FLOSS final report contributes to the theoretical backgroundof the FLOSS project by analysing the Open Source phenomenon, the market forOpen Source software as well as business models for companies based on OpenSource software. It also analyses best practices for the latter. The work is based oncomprehensive literature and online research as well as on several expert talks andpresentations on various conferences and trade fairs (such as the Wizards of OS 2 inBerlin, October 11-13, 2001 or the LinuxWorld in Frankfurt, October 30-Novem-ber 1, 2001).

Organisation of the reportThe report is organised as follows: Section 2.1 gives an introduction into the OpenSource (OS) software phenomenon. The term OS software (OSS) is defined andcompared to several other software distribution concepts. Various OSS license poli-cies are compared and the major OSS products and development projects are intro-duced. Section 2.2 provides the theoretical background for a strategic market analysisof the software market in general. The software value chain for Open Source productsis derived on the basis of traditional software production. For each part of the valuechain, proprietary or commercial software is compared to OS software and free soft-ware.

Chapter 3 provides an overview of the software market and its relation to the IT (in-formation and technology) market. It also segments the software market. Business dy-namics for the software products market, for the software services market, and for theembedded products market are discussed. A basic understanding of market structuresof the software market is necessary to evaluate the business models in Chapter 4. It isalso needed to analyse the impact of OSS on the traditional software market.

Chapter 4 outlines the different OSS business models referring to the market analysisin Chapter 3. For each business model identified the basic business principle, themarket and the critical success factors will be evaluated.

Results went into other parts of FLOSS project

This report formed the basic foundation for other elements within the FLOSSproject. For example, the analysis of different software characteristics went into theconstruction of the user survey and the business models for companies that wantedto establish a business based on Open Source went into the analysis of firms’ OpenSource activities and the resulting policy implications.


1 Introduction

10


11

2 Software and the Open Source phenomenon

2 Software and the Open Source phenomenon

2.1 The Open Source phenomenon

2.1.1 Definition of Open Source software

Availability of source code and price main criteria

There are several forms of software licensing and software distribution. They canmainly be distinguished with respect to two criteria: Availability of source code andprice. Source code is software code written in a higher level programming language.It is different from binary code (code of ones and zeros only, which is readable forcomputer machines) because it reveals the structure and logic of a program. A soft-ware, which is distributed in binary code only, is also called closed source.

Figure 2–1Classification of OSS versus other software

Most important software forms

The most important software forms are:

❑ The classical proprietary/commercial software: This software is typically dis-tributed in binary form only. The source code is not available.

❑ Shareware: Software in this form is typically free for an initial period, but gen-erally after a test period a license has to be bought. The source code is notavailable.

❑ Freeware: For this form, there is no license fee at all, at least not for the free-ware product, but maybe for a complementary product. The source code isnot available.

❑ Open Source Software (OSS): The source code of this software is available.There is commercial OSS (sometimes software that has been closed sourceand the source code was released) and non-commercial software, which can

Source code open

Ber

leco

n Re

sear

ch 2

002

Pric

e fo

r the

use

r

Non-Commercial

OSS

FreewareShareware

CommercialOSS

Proprietary/ Commercial

Software

Yes No

0 (g

rati

s)>0

(non

gra

tis)


2.1 Software and the Open Source phenomenon

12

normally be downloaded from a website. The main characteristic of OSS isthat the users can freely use, modify and redistribute the software. However,there is a rich landscape of OSS licenses, which differ in terms of right forcommercial use and other aspects. They will be examined in section 2.1.3.

Free Software and SharedSource

In addition also the terms “Free Software” and “Shared Source” can be found. „Freesoftware“ is a matter of the users' freedom to run, copy, distribute, study, change andimprove the software. It is thus much closer to „free speech“ than to „free beer“.1

Shared Source is a relatively recent policy by Microsoft to give certain groups of com-panies access to source code of Microsoft product. This does not, however, includethe right to modify the code.

2.1.2 Historic development2

Originally hardware asmajor revenue stream

From the early 1960s to the early 1980s, revenues in computer business were gener-ated through selling and supporting hardware. For every hardware device, a specialoperating system was developed and deployed. The users of these systems were highlyspecialised IT experts. They were the ones primarily responsible for the developmentof additional software.

Unix a basis for thedevelopment of Internet

technologies

Many efforts were dedicated to build an operating system that could be deployed onmultiple hardware platforms. The most prominent example was Unix, which devel-oped at the AT&T Laboratories and was published in 1969. Commercial users hadto pay high license fees for using Unix, whereas academic institutions could use thesoftware for a nominal charge. Consequently, Unix was the basis for the developmentof the Internet technologies. Many of these technologies were developed at universi-ties and computer companies research laboratories, where Unix was deployed. Shar-ing the source code among software developers was commonplace. This tendency wasreinforced by the emergence of computer networks like the Usenet that was startedin 1979 to link the Unix community.

Turnaround in AT&T’slicensing policy.

A critical event in the early 1980s for cooperative software development was the turn-around in AT&T’s licensing policy. Unix became restricted to those who paid for thelicense to use is. Following this first step into the direction of closed source, the hard-ware companies IBM, HP and DEC started to develop proprietary Unix operatingsystems. They imposed “non-disclosure agreements” on the programmers dealingwith the software and recruited many developers for commercial software develop-ment who had formerly contributed to cooperative and shared software development.

Richard Stallman and GNU At that time, the programmer Richard Stallman worked in software development atthe MIT. In 1984, he started a project to develop a free alternative of the Unix oper-ating system. In addition, he established a special license, the GNU (named for Gnu’sNot Unix) license, which was supposed to ensure that the software is indeed free andopen for everyone. In order to support the GNU project, Stallman founded the FreeSoftware Foundation (FSF) in 1985. Although linked often to the Open Sourcemovement, Stallman is a proponent of Free Software, which goes much further in itsdemands.

Nevertheless, the GNU General Public License (GPL, see “Licenses”) is central to theevolution of the Open Source phenomenon and has been used in many important

1. http://www.gnu.org/philosophy/free-sw.html.2. This introduction can only be a very brief overview. To get more into the details see, e.g.,

Rosenberg (2000) and the literature overview.


13


projects. In the GPL, the principle of “Copyleft” is realised: It means that every copyof a program governed by the GPL, even if modified, must be subject to the GPLagain. The licensing principles of the GPL, especially the “viral” effect,3 are not suitedfor use in commercial software development as they make license fee-based revenuemodels impossible.

Antipathy from many businesses towards FSF

The FSF’s philosophy behind software development provided great motivation forthe Free Software community. But it also resulted in antipathy from many businesseswhich partly remains until today. The most prominent debate over the implicationsof Open Source Software, especially the GPL, and its effects on innovation takes placebetween Microsoft and Free/Open Source Software advocates, although such discus-sions are commonplace in more prosaic settings as well.

LinuxIn the early 1990s, along with the increasing use of the Internet and the success of theWorld Wide Web, many new Open Source projects emerged. The most prominentexample is Linux. Linux is a Unix-like operating system targeted to run on a personalcomputer. It was developed by the Finnish computer science student Linus Torvaldswho used the GNU software tools. In 1991, he released the code of an experimentalversion under the GPL to a newsgroup and asked for comments and improvements.Within the last decade, Linux developed into a powerful operating system. Theproject shows characteristics that are typical for successful Open Source Software de-velopment over the Internet.

Cathedral vs. BazaarEric Raymond, another central OSS developer and advocate, describes OSS develop-ment coordination as “Bazaar style,” opposed to the “Cathedral” approach taken inclassical software development, where development is organised in a more hierarchic,top-down and planned way. Linux has a modular structure, so individuals or groupsof developers can focus on one part of the program. The principle of “Release often,release early” in combination with a constant peer-reviewing process (“Given a thou-sand eyes all bugs are shallow”) is also opposed to commercial software development.(See also section 2.2.3 on software development).

OSI approach to software licensing

Linux was used increasingly in combination with the GNU tools. Because the oper-ating system is central to IT infrastructure, it eventually became relevant for businessuse. In 1997, the Open Source Initiative (OSI) was founded in order to establish amore pragmatic approach to software licensing. The OSI was based on the “DebianFree Software Guidelines,” which had been published in 1995. The central people forthis development were Eric Raymond and Bruce Perens. Their aim was to promoteOSS in commercial use because they believed that both the Free/Open Source com-munity and the business world could benefit from wider OSS dissemination.

Open Source Definition by OSI

The OSI developed the Open Source Definition (OSD). The definition is not a li-cense itself, but a guideline and trademark for OSS software licenses other than theGPL. Licenses according to the OSD guarantee several freedoms to software users,including commercial users. The “viral” effect of the GPL is not a requirement forOSD-approved licenses. In order to raise acceptance of OSS in the business world,the term Open Source Software instead of Free Software was established and widelyaccepted.

Rise in attention paid to OS projects in 1990s

The 1990s experienced a significant rise in attention paid to Open Source projects.Many companies from the IT industry began to support the projects. IBM, for exam-ple, supports a variety of Open Source projects. In 1998, Netscape was the first prom-inent company to release a proprietary software product as Open Source software.

3. “Viral” means that every derived software or software linked to a software published under theGPL has to be licensed under conditions that are compatible to the GPL.



14

Along with the Internet boom in the late 1990s, many investments were made inbusiness models based on OSS. For example, Red Hat and VA Linux reached an in-credibly high market capitalisation. Today, many of these companies are strugglingor have changed their business models. But the development of OSS has not ceased.There are thousands of different ongoing projects. The most important projects willbe described in section 2.1.4.

2.1.3 Open Source licensing

Most important part of thecopyright is licensing

The terms “open source” or “free software” are not specific enough to describe a par-ticular software license. In general, the most important part of the copyright in thesoftware business regards the software’s copying, distribution and preparation of de-rived works (Rosenberg, 2000, 89). The Open Source licenses have two things incommon: The right to earn license fees is typically waived and the condition that thesource code is made available is incorporated.

Copyleft The major distinction of Open Source licenses is the principle of “Copyleft”. “Copy-lefting” a software product means to impose the restriction on the user that derivativeworks have to be released under the same license (i.e., as Open Source) again. Themost prominent and strongest case of “Copyleft” is the GPL. The “Copyleft” is notincluded in the opposing BSD-style licenses.

The restrictions and freedoms in an OSS license have to be balanced carefully whencrafting a software license to both satisfy the customer’s needs and the copyright own-er’s intentions (e.g. maximize the software’s use for all users, including commercialorganisations, or to ensure that the software remains “free” in the sense of the FSF’sphilosophy). There have been many attempts to define the ideal Open Source license,resulting in many different licenses in the open source area. In the following, we willshortly describe the most important licensing models in order to define Open Sourcemore precisely.

Public Domain

Public domain abandons allcopyrights

Releasing software to the “Public Domain” means to abandon all copyrights. ThePublic Domain principle is only applicable within some legal environments such asthe U.S. In Germany, publishing software according to the “Public Domain” is notconform to German right (Urheberrecht). In the U.S., Public Domain software islargely developed with government support at universities or research institutes. It isavailable for every US citizen without any restriction. Taking the software into thecommercial domain is allowed.

Shareware

Shareware make softwareavailable in binary form for

small license fees

The intention behind shareware is to make a software programs available to as manypeople as possible. Shareware is distributed in binary form only. Most copyright own-ers impose a small fee for the license that normally has to be paid after a certain testphase.

Shareware evangelists argue that software producers want to be compensated for theirwork and a certain degree of “fair use” is needed. For freeware developers, it is sup-posedly even more difficult than for Microsoft to enforce their rights. Sharewareproducts often have a built-in mechanism that dramatically reduces the comfort ofusage after the trial period. This mechanism is intended to increase the willingness ofthe user to pay for the license.


15


Freeware

Freeware is freely available in binary form

Freeware is distributed in binary form without imposing license fees for usage. It ispossible to grant the right to use the software exclusively to certain users, e.g. privateor non-commercial users. Freeware often is part of a marketing strategy to promotecomplementary products. For example, Microsoft released the Internet Explorer asfreeware to gain market share.

GNU Public License (GPL)

GPL most widely used Open Source Software license

The GPL is the most widely used Open Source Software license. The most prominentexamples are the GNU project and Linux. The GPL was created by Richard Stallmanand is representing the Free Software Foundation’s philosophy.

There is no restriction on copying and distribution, but some conditions have to bemet: The source code must be easily available to the user, the GNU GPL has to beenclosed with distributed software and interactive programs have to display the no-tices when started. Modifications are allowed as long as the author states what thechange is, when the code was written and by whom. Derivative works are permittedbut have again to be published under the GPL (“Copylefted”). The so-called “viral”effect arises because software incorporating GPL software must also be licensed undera GPL-compatible license. Software under another license can be “contaminated.”The very strong “Copyleft” makes GPL not very business-friendly because any soft-ware company would have to reveal their software source code if they used (in thesense of included) parts of GPL software to develop it.

GNU Lesser General Public License (LGPL)

LGPL less restrictiveBecause the GPL is so restrictive to commercial use, the FSF developed the LGPL.The LGPL allows commercial software to use libraries without being “contaminated”by the GPL. One main rationale was that a superior Free Software product could notbe disseminated widely enough to become a standard and compete with commercialproducts. The first software issued under the LGPL were the GNU C libraries.

Mozilla Public License (MPL)

MPL license for Netscape browser source code

The MPL is the license under which Netscape released the source code of its Netscapebrowser client, Mozilla. The MPL imposes a kind of a “Copyleft” on the usage ofMPL-software. The main difference to the GPL is that software under MPL can beincorporated into software products that can be licensed without “contaminating”the software. Therefore, the basics of the MPL are similar to the LGPL. Similar li-censes are the IBM Public License or the Sun Public License. All of these licenses areOSI-approved.

MIT License and BSD License

MIT and BSD licenseThe BSD license grants the right to obtain a copy of the software including documen-tation materials available free of charge to everyone, as well as the right for trading theproduct commercially. Redistribution and use in source and binary forms is allowed.The licensing terms also hold for copies and ported versions of the software.

Part of the original BSD License was that the copyright owner’s and the project’s con-tributors names must not be used for promotion of a derived product without writtenagreement. But in the FreeBSD License, that condition is not mentioned anymorefor practical reasons.



16

Table 2–1Licenses overview

2.1.4 Main Open Source products and development projects

Apache4

Development started atNCSA

Until 1995, the National Center for Supercomputer Applications (NCSA) providedthe most widely used web server. The software could be copied and modified free ofcharge provided that NCSA would be notified about such derivative work. The usersturned over software patches to the NSCA where they were incorporated into theNCSA software. In 1995, many NCSA developers changed to Netscape, so the main-ly professional users were not satisfied with the software’s quality anymore and putup a mailing list furthering development of the NCSA web server, then calledApache.

The Apache Software Foundation (ASF) was founded in 1999 and is now responsiblefor Apache development. Apache is known as a very good example for successful col-legial management of the open source software development process. Brian Behlen-dorf as the community member responsible for public relations and involved in theproject from the beginnings is strongly connected with Apache.

Apache dominates market The Apache web server has dominated its market since 1996. Today, its market shareis around 60 percent (Netcraft, 2001), leaving Microsoft IIS behind with around 30percent. The Apache web server is included in many software solutions (e.g. IBM’sWebSphere).

DNS and Bind5

BIND is foundation ofdomain name system

The Berkeley Internet Name Daemon (BIND), delivering the Domain Name System(DNS), is not very well known among general IT users. Nevertheless, as a programfor turning host names into IP addresses it is a very important component of the In-ternet infrastructure. BIND is included in all UNIX systems and in many other sys-tems, as well as being the de facto Internet standard for its functionality.

BIND was initially developed by Paul Mockapetris in 1984 and is now under theleadership of Paul Vixie from the Internet Software Consortium (ISC). The ISC was

Software license Available at no cost

Distribu-tion allo-

wed

No usage restric-tions

Source code fre-ely avai-

lable

Source code mo-dification allowed

Derived work must

be free again

Linking with prop-

rietary software allowed

Public Domain X X X X X X

Shareware (X)1 X

Freeware X X X

GPL X X X X X X

LGPL X X X X X X X

MPL X X X X X X X

BSD-License X X X X X X

1) Shareware is gratis for a trial period only

4. Cf. www.apache.org; Lerner and Tirole, 2000; O’Reilly, 1999.5. Cf. www.isc.org/products/BIND/; www.isc.org; O’Reilly, 1999.


17


founded in 1993 from Rick Adams with a donation from UUNET and is now sup-ported by major players from the software industry (BIND Version 9 has been un-derwritten by Sun, HP, Compaq, IBM and SGI).

Free BSD6

FreeBSD: high-quality software without getting involved with money

In 1993, the first version of FreeBSD was released. It was initially built on the Berke-ley Software Distribution (a free Unix system), and provides an Open Source Unixoperating system. It is licensed to a large extent under a variant of the BSD license,imposing little restrictions on both commercial and non-profit users. NetBSD andOpenBSD are other Open Source projects built on the BSD, while FreeBSD is saidto be the most popular variant. FreeBSD release engineer Jordan K. Hubbard statedthat FreeBSD’s goal was to produce “commercial” software in terms of functionalityand quality without getting involved with money.

FreeBSD partly foundation of MacOS X

When FreeBSD was looking for a distribution channel, Walnut Creek became themajor distributor and supported FreeBSD financially and through IT infrastructureequipment. One of the core developers, Jordan K. Hubbard, is now employed by Ap-ple, which is strengthening the ties between Apple and the FreeBSD community. TheApple operating system Mac OS X is largely based on FreeBSD. WindRiver, ownerof the FreeBSD trademark and distributor of the commercial BSD/OS stopped its fi-nancial engagement with FreeBSD, releasing the FreeBSD developers employed atWind River and stopping technical support.

Gimp7

Gimp: OS graphics softwareThe Gimp (GNU Image Manipulation Program) is a graphics software often calledthe “free Photoshop” (a well known commercial graphics software from Adobe). Itwas developed by Peter Mattis and Spencer Kimball for Unix systems and is now partof most Linux distributions.

GNOME8

GNOME: Open Source desktop environment

GNOME (GNU´s Network Object Model Environment) is competing with KDEin the marketplace of Open Source desktop environments for Linux and other Unix-like operating systems. Gnome is using the GTK (GNU GUI toolkit) and version 1.0was released in March 1999 with Miguel de Icaza as the chief developer.

In May 2001, Eazel, a company developing a GNOME distribution that would bringmore user-friendliness to GNOME, closed its doors. Nevertheless, Miguel de Icazaas the cofounder of Ximian Inc. is also pursuing the goal to bring Linux to the desk-top. But differing from Eazel, Ximian is targeting business users, for example for useat point-of-sales terminals. In August 2001 Dell stopped delivery of its PCs with theLinux operating system because of its user-unfriendliness. Nevertheless, theGNOME project is supported by Sun, IBM or HP and has a strong community (ashas KDE) making rapid evolution possible.

6. Cf. Yager, 2001; Michaelis, 2001; www.freebsd.org; O’Reilly 1999.7. Cf. www.gimp.org; O’Reilly, 1999.8. Cf. www.gnome.org; www.ximian.com; de Icaza, 2001; Hall, 2001; O’Reilly 1999.



18

GNU9

GNU project aimed at freeUnix

GNU is an acronym for “Gnu is not Unix”, expressing the starting point of thisproject very well. Richard Stallman from MIT started the GNU project in 1984 withthe aim to build a freely available Unix-like operating system. When he quit MIT tofocus on the GNU project, MIT continued supporting him with access to their in-formation technology resources. The GNU project has not yet succeeded at buildingits own competitive OS kernel (Hurd), instead the combination of GNU and Linuxis very successful. Very important determinants of the GNU project’s evolution is theFree Software Foundation’s philosophy and the GNU General Public License (GPL).The Free Software Foundation (FSF) was founded in 1985 and is financed by dona-tions and fees for manuals and free software distributed on physical media.

The GNU project today is involved in many software projects related to their Unixoperating system clone and the related tools that would offer a choice to commerciallydistributed software. Very important software products are the GNU C Compiler(GCC, the de facto standard compiler for Linux programs written in C and C++), theGNU Privacy Guard (GnuPG), GNU Emacs (a text editor) or GNOME.

KDE10

KDE: graphical userinterface for Unix systems

KDE, the “K desktop environment”, was founded in 1996. The aim was to developa graphical user interface (GUI) for Unix systems designed to put an end to the het-erogeneous and user-unfriendly Unix GUIs existing so far. It is now available in ver-sion 3.0.

KDE is GPL compatible and the KDE libraries are available under the LGPL, makingcommercial software development for the KDE desktop possible. All KDE applica-tions and the important tools KDE is built on are licensed under the GPL.

A complementary product, Koffice, is developed as an open source office applicationsuite for KDE, making KDE and Koffice one of the primary potential competitorsfor Windows in the desktop market.

Linux11

Linux today the best-knowOSS project

The Linux project is tied to one person: Linus Torvalds. As a computer science stu-dent, he wanted to run a Unix-like operating system on his personal computer. Asthere was no satisfying software available, he started developing his own OS from“Minix”, a Unix operating system for educational use created by Prof. Tanenbaum.When the software reached a certain degree of maturity, he posted the code on a dis-cussion group asking for comments and possible improvements. Today, Linux isprobably the best-known OSS project. As Linux is a Unix-like operating system ker-nel (providing the basic functionality of an operating system), it is deployed with theGNU tools.

The founder Linus Torvalds is still in the centre of Linux development. The Linuxcommunity consists of thousands of developers coordinated in the “bazaar style”. Butnevertheless, there are some hierarchic structures with Torvalds on top and so-called“Lieutenants” managing sub-projects.

9. Cf. www.gnu.org; O’Reilly, 1999; Freyermuth, 2001a, 2001b.10. Cf. www.kde.org; Eduak, 2001; O’Reilly, 1999.11. Cf. Computerwoche Spezial, 2000; O’Reilly, 1999; Rosenberg, 2000; Schmitz, 2001; Freyer-

muth, 2001a, 2001b.


19


Many organisations make up Linux ecosystem

There are many organisations involved: Mostly commercial players like the distribu-tors (Red Hat, SuSE, Mandrake Soft) and some major hardware and software com-panies (IBM, HP) porting their software and hardware to Linux have a direct interestin Linux. Along with employing developers working on Linux they provide infra-structure resources (for example the Open Source Development Lab sponsored byIBM and HP among others). The Linux Standard Base (LSB) as a consortium ofcommercial and non-profit organisations is trying to establish common standards forLinux development (to prevent the emergence of diversity that happened to Unix inthe 80/90s).

Linux released under GPLLinux was released under the GPL. The GNU/Linux system grew in importancemainly in the server market, partly as a substitute of commercial low-end Unix sys-tems and partly as a competitor of Windows NT. As a desktop operating system, themarket share is low, but complementary developments focusing on user friendliness(e.g. GNOME, KDE, OpenOffice) may evolve into competitive Linux solutions forconsumers. The main characteristic is that Linux is a very mature and in many waystechnically superior operating system which has not forked as happened to Unix(with its existing commercial products like Sun’s Solaris and the several free versions,namely FreeBSD, NetBSD, OpenBSD). At the moment, that is one major issue infurther Linux development targeted by the LSB, as the different distributions add dif-fering functionalities to the kernel, resulting in interoperability problems.

Mozilla12

Mozilla: Open Source browser project initiated by Netscape

Mozilla is the name of an Open Source browser project initiated by Netscape. WhenMicrosoft began offering the Internet Explorer for free and improving its perform-ance, Netscape released the code not protected by third party copyright law for itsbrowser. The license created was the Mozilla Public License.

When releasing the code, the hope was to leverage the power of the open source de-velopment process. According to some outsiders’ estimations, initially there was nolarge developer community and Netscape developers accounted for most of the con-tributions to the code. Apart from the reason that Netscape could have retained toomuch control over the Mozilla project, the code base is very large and the programvery complex. That makes the program difficult to understand and adds to the prob-lems of making Mozilla a successful OSS project. Ideally, there should be a modularsoftware design (as is the case with Linux) for being able to share the work. Only in2002 Mozilla released its 1.0 version.

MySQL13

Relational database MySQLMySQL is a relational database server, initially developed in 1994. The Swedish com-pany TcX Dataconsult AB, that later took on the name MySQL AB, published thesoftware under the GPL (and parts of it under the LGPL) in 2000. It also offers ver-sions under a license that enables the owner to use MySQL in commercial solutions.MySQL AB, a Swedish privately held company, which is financed with venture cap-ital since July 2001, is the owner of the copyright. Major companies have deployedMySQL, e.g. Motorola or Yahoo!.

12. Cf. www.mozilla.org; Rosenberg, 2000; Miller, 2001.13. Cf. www.mysql.com; www.nusphere.com; Jaeger, 2001a; Wayner, 2001; Dyck, 2001.



20

NuSphere is a company of Progress Software and offered a MySQL distributionalong with solutions that integrate MySQL with Apache and PHP. In July 2001,MySQL AB accused NuSphere in Massachsetts/USA of violating the GPL, the firsttime a violation of the GPL was brought to court.

Perl14

Scripting language Perl Perl was developed in 1987 by Larry Wall. First being a tool for searching, manipu-lating and printing texts, the scripting language Perl evolved into a network and sys-tems administration tool. The CGI-programming functionality made it the “glue” ofthe Internet, especially suitable for dynamic web pages.

Wall is further developing Perl with around 100 other developers while employed atO’Reilly. ActiveState Tool, for example, is offering professional tools for Perl, and theComprehensive Perl Archive Network (CPAN) gives access to hundreds of Perl mod-ules.

PostgreSQL15

Relational databasePostgreSQL

PostgreSQL is an object-relational database server. Its roots are at UC Berkeley; in1996 a team took the existing code and developed it into an open source SQL-data-base. It is mainly deployed in private use and projects not involving mission-criticalbusiness operations. Other reasons keeping PostgreSQL from spreading may be its in-complete documentation and the lack of professional support, as have many OSSprojects in the beginning.

In September 2001 Great Bridge, the company employing 3 of the 6 core Post-greSQL developers ceased operations. Their main goal was to offer professional serv-ices to businesses. Nevertheless, indicating PostgreSQL’s quality, Red Hat offers itsown database product built on PostgreSQL 7.1.

Python16

Scripting languagePyhthon

Python is a scripting language first released in 1991 by Guido van Rossum. It evolvedquickly into a powerful object oriented and interpreted programming language, oftencompared to Perl or Tcl. The product is offered in a version completely integratedwith Java (Jpython), enabling it to run on every computer with the Java Virtual Ma-chine installed.

Van Rossum is working for Zope Corporation as the director of the Python Labs andis the leader of the project. The PSF (Python Software Foundation) was created inMarch 2001 “with the specific goal to own the intellectual property that comprisesthe Python programming language.” Major PSF sponsors are Zope Corporation andActive State.

14. Cf. www.perl.org; O’Reilly, 1999; Lerner and Tirole, 2000.15. Cf. www.postgresql.org; Wayner, 2001.16. Cf. www.python.org; www.zope.org; O’Reilly, 1999.



Samba17

Samba for print and file server connection

In 1993, Samba was developed by Andrew Tridgell (who is still leading the project)from the Australian National University as a Windows file server and print server forUnix platforms. The current version, 2.2.x, implements SMB and CIFS (two net-work protocols) and adds to the comfort of replacing Windows servers with, e.g.,Linux servers. Samba is included in most of the Linux distributions.

There have been some tests running Samba as a file or print server against Windowsservers. SGI was said to offer the fastest Windows NT fileserver using Samba. And inNovember 2001, PC Magazine tested Windows 2000 against Samba/Linux as a printserver, the latter outperformed Windows on both low and better quality systems.

Sendmail18

Sendmail handles majority of mail traffic on the Internet

Sendmail was developed in 1981 by Eric Allman as a Mail Transfer Agent (MTA).At that time he worked at UC Berkely and wrote sendmail to exchange mail betweenthe university’s network and the Arpanet. (Freyermuth, 2001a, 7). From the begin-ning, Sendmail was focused on openness regarding differing mail protocols, on therouting functionality (not as many other programs getting involved with the frontend or mail delivery) and on a flexible configuration file.

Allman is still central to sendmail development. In 1993, Allman had to rewrite theprogram as there existed many forks, and succeeded in reuniting the communities.Complementing the free basic sendmail program, he and Greg Olson founded Send-mail, Inc. in 1997. Sendmail, Inc. offers commercial versions of sendmail, addingmanaging tools and security solutions. Sendmail dominates its market, reaching attimes a market share of 75 to 80 percent of mail delivered via sendmail.

StarOffice/OpenOffice.org19

StarOffice suite as competitor to Microsoft Office

The StarOffice suite is a Sun Microsystems product since Sun took over Star Divisiontwo years ago. It is competing in the office suite market dominated by Microsoft Of-fice. Corel and Lotus Smart Suite are other competitors. A year ago, Sun made theStarOffice code available as the “OpenOffice.org” project. The basic software can beused and commercially exploited by everyone, competitive advantage is gainedthrough adding proprietary extensions, e.g. by Sun in its StarOffice product (whichhas to be licensed for a fee since May 2002) with a proprietary spell checker.

Tcl/Tk20

Tcl/Tk for developing graphical user interfaces

Tcl/Tk was developed by John Ousterhout while he was a Professor at UC Berkeley.Tcl is a scripting language comparable to Perl and Python, Tk is a toolkit for devel-oping graphical user interfaces (GUI). Initially developed for Unix systems, Tcl/Tknow is also supporting Microsoft Windows and Apple Macintosh platforms. Ouster-hout was working with Sun until he founded his own company “Scriptics” (with asimilar business model as sendmail, Inc.) to further the development of Tcl/Tk.

17. Cf. www.samba.org; Kaven, 2001; O’Reilly, 1999.18. Cf. www.sendmail.org; www.sendmail.com; O’Reilly, 1999; Lerner and Tirole, 2000.19. Cf. www.openoffice.org; www.sun.com/staroffice/; Eduak, 2001.20. Cf. www.scriptics.com; O’Reilly, 1999.


2.2 Software and the Open Source phenomenon22

Zope21

Zope: Open Sourceapplication server

Zope is one of the projects based on initially proprietary developed software. In 1995,Digital Creations was founded by Rob Page and Paul Everitt. They were developingsoftware for newspapers to manage advertising. Later on, they had two main softwareproducts: the free, open source toolkit Bobo and the commercial web applicationplatform Principia. They did not have the financial means to enter the establishedweb application server market (with competitors as IBM Websphere or BEA Web-logic), so in 1998, they adopted the typical Open Source Business model shiftingtheir business from license revenues to a services company’s business model.

By releasing Principia as OSS, they hoped to push marketing and build up a usercommunity that would help to improve the product. Today, Zope is renowned forbeing a competitive alternative in the application server market, especially for contentmanagement and portals. In Germany, several companies offering Zope servicesfounded the “Eurozope Association.” And Digital Creations even changed their nameto Zope Corporation to express the importance the project has for the company.

They are also engaged in the Python project as a sponsor to the Python SoftwareFoundation and employer of Guido van Rossum, leader of the core developmentgroup working at Zope, the PythonLabs.

2.2 Software characteristics and the software value chain

2.2.1 Software characteristics

Software features ingeneral

Following the introduction on Open Source software, we will now take a closer lookat software and its features in general. Firstly, software is immaterial, which makes itdifferent from physical goods. Secondly, users of software rarely buy only the prod-uct, instead they buy the service related to the software. Enterprise users normally buysolutions, a combination of software, hardware and services. For our analysis, we needthe following basic differentiation:

Software defined to consistof product and service

For easier discussion we will define Software as consisting of two substantially differ-ent parts, which will be called product and service. The product is the license that issold to use the software. This license can be unique (customised software product) orit can be duplicated as many times as possible (standardised software product). Theservices surrounding software products range from consulting, implementation, sup-port, and training to application management. Therefore, we will be using the termsoftware product (not software itself) and software services.

Additionally, one can differentiate between individual and standard products andservices. This gives the following matrix:

21. Cf. www.zope.org; Osterberg, 2001; Rosenberg, 2000.



Figure 2–2Matrix of products and services

The software product is determined by several characteristics that make it differentfrom physical products and, therefore, influence the way the quality of software prod-ucts can be determined.

Software product characteristics

The product characteristics can be summarised as follows (Balzert, 1996, 26). Soft-ware ...

❑ Is an non-material product with no physical limits.❑ Cannot be worn out and does not physically deteriorate.❑ Is a product of intellectual property. This results in high development costs

and low per-item costs for standardised products. The incremental costs foradditional sales are negligible. Therefore, volume is very important for sales.

❑ Is aging (better hardware allows better software). This aspect is becoming lessimportant over time.

❑ Is difficult to measure in terms of physical product measures. Either technicalor financial equivalents have to be found to measure the value of software.

2.2.2 The software value chain: overview

Figure 2–3Software value chain

Value chain with product-related and service-related steps

In the software value chain, we see again product-related and service-related steps.Additionally, there are marketing and distribution steps. Here, value is created inform of information about the products.

The value chain consists of three major parts:

❑ Production/Programming,❑ Marketing/Sales (Distribution), and❑ Services.

Products Services

Standardised

Customised

StandardisedProducts

StandardisedServices

Individual Productsand Services(Solutions) B

erle

con

Rese

arch

200

2

SoftwareDevelopment

Software Documen-

tation

Software Packaging

Ber

leco

n Re

sear

ch 2

002

Software Value Chain

Marketing and Sales

Consulting Implemen-

tation/ Integration

Training Support Application

Management

Production/Programming Services



Differences betweenstandard and individual

software

It is instructive to note, once again, the differences in the value chain for standardsoftware and for individual software. Although the same steps need to be performedto create similar value in the form of a software product or solution, the order of thesteps is different. It is not surprising that, for individual solutions, the first step is con-sulting and the definition of requirements for any individual customers. For standardsolutions, the first step is the software development. (Assuming that market researchhas been conducted before.)

In the following section, the steps of the value chain will be analysed separately. Wewill describe first what the essence of each step is. Furthermore, we will point out ma-jor differences between proprietary software and OSS.

Figure 2–4Software value chain for

standard software versusindividual software

2.2.3 Programming and production

Software development

Definition of softwaredevelopment

Software development is the writing of technical code in order to initiate and controlthe functionality of machines, especially computers, in a certain way and so that theymeet certain requirements. Software development can result in either standard or in-dividual software products. It includes the development of productivity tools. Pro-ductivity tools are technical codes (such as programming languages or compilers)which allow to change and control other technical codes more easily.

Trends of softwaredevelopment

Software development is characterised by the following trends (Balzert, 1996, 27).The indicators given in parentheses can be used to measure the relevance and impor-tance of these trends for certain software.

❑ Increasing importance of software versus hardware products (indicator: rela-tion hardware products sales vs. software products sales in the IT market).

❑ Increasing importance of software-related services versus software products(indicator: relation sales of software products vs. software-related services).

❑ Increasing complexity (indicator: number of person-years or number ofsource code lines, measured for consecutive releases of a software product).

❑ Increasing quality requirements (indicator: number of defects per 1,000 linesof source code).

❑ Existence of a trend from individual software products towards standard soft-ware products within a certain product line. This trend corresponds to the

Software Development

Software Documen-

tation

Software Packaging

Marketing and Sales


tation/ Integration


Management

Production/Programming Services

Standard Software (one-to-many)

Consulting Software Develop-

ment

Software Documen-

tation

Software Packaging

Implemen-tation/

Integration Training Support

Application Management

Services (incl. Production/Programming)

Individual Software (one-to-one)



classical cycle of development for technological products. At first, a specialsoftware product is developed to address an individual problem, resulting inindividual software. Later, as the demand for similar problem solutions in-creases, the software becomes standardised. (indicator: relation of customisedsoftware sales to standard software sales).

❑ Increasing burden of already existing software (adaptation of old software ver-sions to new components accounts for two thirds of software developmentcosts, trend increasing).

Software development There are basically two principles of developing software (Raymond, 1997):

Cathedral…❑ Cathedral-principle: This principle is based on centralised planning and exe-cution and describes the way of traditional proprietary software development.Software development is organised top-down within a strong organisationalhierarchy. The name of the principle refers to the building of cathedrals in themedieval age.

…and bazaar❑ Bazaar-principle: This principle is based on decentralised planning and execu-tion and explains the way of much Open Source software development. Thedevelopment is organised in a networked manner. A community of developersworks on parts of the software and on improving it. The patches are broughttogether via the Internet on a central website. The name of the principle refersto an oriental bazaar where anyone can trade anything and anyone can partic-ipate.

Cooking-pot modelA principle which goes even further than the bazaar principle is the cooking pot mod-el (Ghosh, 1998) with following argumentation: On a bazaar, traded goods eventu-ally change their possessor but their value is not changing. In the cooking pot, instead,single ingredients become melted and result in something new, which is valued higherthan the mere sum of the ingredients.

Development models often blended

In reality, these principles of development are sometimes blended. For example, asoftware product is developed in the cathedral-principle until its first release. The firstrelease is further developed in the bazaar-principle. Users can add modules and func-tionalities.

Critical success factorsThere are several critical success factors on this part of the value chain:

❑ Functionality requirements must be met, the software must perform in the de-sired way.

❑ High software quality must be achieved.❑ Time-to-market must be short because of adaptability to new hardware envi-

ronments. However, this factor is becoming less important because hardwaredevelopment and substitution seems to slow down in some product lines.

❑ Development costs must be minimised.Differences between proprietary software and OSS

There are different development models and a different developing process for pro-prietary software and OSS. Proprietary software allows for the centralised definitionand direction of functional requirements. OSS develops in whatever direction indi-viduals want it to. Often reality is less distinct, as also commercial entities can adopta bazaar-like development model at least to some extent. Also Open Source develop-ment is often more hierarchical (e.g. with “Lieutenants” deciding about which ele-ments to include) than in the pure bazaar model.

It is sometimes hypothesised that debugging seems to be faster and better with OSSbecause an entire community is addressing problems. In numbers, OSS should there-fore be of higher quality as measured by number of defects per 1,000 lines of sourcecode.



Another often stated difference is that both sorts of software seem to differ in the pri-orities of development. Proprietary software tends to emphasize usability. OSS tendsto emphasize stability and functionality because it is aimed at a different audience.

Major difference between both is that the source code of proprietary software is notopen for the user. Therefore, adaptations of software to special hardware conditionsor functionality conditions can only be made in the form of customisation or cannotbe made at all. For OSS, the source code is available and gives many more opportu-nities to further develop the software according to special requirements resulting inbetter performance.

Software documentation

Definition of softwaredocumentation

Software documentation is the aggregation and editing of information related to theapplication and use of software products. There are three kinds of documentation:Firstly, the software itself is a form of documentation. If the source code is available,software developers can use the code to get information about the product. Secondly,there is electronic documentation explaining the software code within the software.Thirdly, there is edited documentation, usually in paper form, explaining the func-tioning of software.

Differences betweenproprietary software and

OSS

Proprietary software is usually sold with documentation. Documentation is part ofthe developing process at some late stage. In comparison, OSS is normally not a fin-ished (box-)product – therefore, documentation is not always sufficient for the user.Documentation production and sales are part of the business model of some compa-nies related to OSS, such as Red Hat (electronic documentation) and the O’Reillypublishing company (paper documentation).

Software packaging

Definition of softwarepackaging

Software packaging is the aggregation, integration and optimisation of software prod-ucts or components. (It can include the debugging of the new integrated softwareproduct.) Simple aggregation of different software components (e. g. on a CD-ROMor on a website) is not part of software packaging as it adds no value on the productlevel. It adds value on the distribution level and, therefore, part of the marketing/salesvalue chain element. Hence, software packaging assumes that there is value added tothe package and the packaged software product is different from the single parts add-ed.


OSS

Proprietary software usually is sold in a packaged way. Packaging is part of the devel-oping process at some late stage. For OSS, packaging is an important part of manybusiness models based on OSS. As the original OSS usually only exists in form ofmany different development projects on the web, the so-called distributors put thecurrent versions of the software together, optimise them and then sell the package.

2.2.4 Marketing and distribution

Goal: optimise instrumentsof marketing mix

The general goal of marketing and sales activities is to optimise the combination ofinstruments within the marketing mix (product, price, distribution, advertising) inorder to create an awareness within the target group (potential customers) for theunique selling proposition of the product. Eventually, this awareness is aimed to re-sult in the purchase of the product.



Marketing and market structures for standardised products and services (see chapter3) are very different from marketing and market structures for individual productsand services (and solutions). As marketing and distribution is more important andmore complex for standardised products and services, we will focus our analysis onthese. There are several critical success factors on this part of the value chain:

Moore’s lawAn often cited success factor in the standardised products market is called Moore’sLaw: According to this “law”, one has to reach 40 percent market share within a cer-tain niche within 12 to 18 months. At this level of success – assuming the closest com-petitor is far behind – word of mouth starts spreading the message that one is themarket leader. Once that happens, one’s market share is expected to increase above50 percent within the following 12 months. Moore’s law is based on the assumptionthat successful products tend to become even more successful whereas less successfulproducts tend to fall even further behind. Network effects can explain this.

Differences between standardised proprietary software and OSS

Difference between the two sorts of software in marketing and distribution issues aresimilar to the differences in documentation and packaging: Marketing of the softwareproduct itself is normally done by the company that develops the software. For OSS,however, this is different. The open source software product itself, i.e. the license touse the software, is usually not promoted separately (at least not with traditional mar-keting instruments like advertising). Only packaged distributions, complementaryproducts and OSS services are promoted.

There are more differences to be found in the four instruments of the marketing mix.

ProductProduct-related issues regard things like quality, time-to-market, and thus manycharacteristics directly resulting from the development process. The product dimen-sion, however, also includes quality issues of the service dimension because, in thesoftware value chain, not only the products but also the services are sold.

PriceThere is a major difference between proprietary software and OSS. (See also Chapter2.2.1 for the difference between Freeware, Shareware and OSS.) Proprietary softwarecan be sold (or rather the right to use the software in form of a license). Therefore,different pricing strategies make sense. OSS can be sold as well, but in form of pack-aged OSS. The price that OSS distributors can get for their software packages is nor-mally significantly lower than the price that competitors offering proprietary softwaredemand. This is due to the fact that less value is being added by packaging than bysoftware development from scratch. Pricing itself is a delicate issue for OSS becausethe model to sell OSS does not really comply with the philosophy of many OSS com-munity members.

DistributionDistribution or sales is part of the marketing mix. Its primary goal is to physically de-liver the product to the right place, where it can be accessed, used, or bought by thecustomer. Distribution of proprietary software is normally done by the software com-pany that develops the software product. Alternatively, distribution is done by sepa-rate distributors or VARs (value-added resellers). VARs operate either exclusively orindependently, sell combinations of software and hardware and act as a sales channelfor software product companies.

Moreover, aggregation is part of distribution. Aggregation is the joint distribution ofsoftware components without software product changes. No additional features areadded. (Aggregation is not part of the software packaging value chain, but it is partof distribution.) Aggregation for proprietary software products is part of the develop-ment process. Aggregation for OSS, instead, is done by distributors and forms thebase for a separate OSS business model.



Advertising and Branding In the mass market, advertising has an important function. For OSS, and for operat-ing systems (Linux) in particular, branding is even more important because many us-ers (especially the ones without deeper software knowledge) choose their software onpopularity. In addition, the OSS distributors and other OSS-related companies havea vital interest in promoting the use of OSS in general. This interest is supported bythe various OSS organisations such as the Open Source Initiative or the Free SoftwareFoundation.

2.2.5 Services

Consulting

Definition of consulting Consulting, in the context of software business and IT-technology, is a software-re-lated service. Consulting includes the following steps:

❑ Analysis of the current situation.❑ Definition of requirements (Conception of functional and technical require-

ments).❑ Selection of software solutions.

Additionally, a normal IT-consulting project would include the steps of implemen-tation, integration, tests and training of the users. All these steps will be discussed sep-arately in the sections below.


OSS

There is no difference in the consulting process itself and in the type of company thatoffers the consulting service. Although there are many companies that specialize inOSS products and offer special knowledge in this category, their consulting processdoes not substantially differ from companies which are focused on proprietary soft-ware.

Implementation and integration

Definition ofimplementation and

integration

Implementation includes the following steps:

❑ Installation (to make a software run on an existing basis of software or hard-ware).

❑ Configuration (to adjust the software to customer needs within the givenframework without changing the source code of the software).

❑ Customisation (to adjust the software to customer needs, changing of thesource code is necessary, additional systems integration can be necessary).

Integration can be in form of systems integration, networking integration and appli-cation integration. Systems integration can be part of customisation if it is necessaryfor the functionality of the software to integrate it in existing systems. Furthermore,integration can be a separate part of the value chain if the primary goal is not to im-plement new software but to integrate existing systems, networks or applications.


OSS

There is basically no difference if the source code and the tools are available. A generalargument in favour of OSS is that the open source code enables customisation, whichis not possible with closed source code. Hence, OSS offers unlimited opportunitiesfor customising a software product.



Training

Definition of trainingTraining on how to use software can be given either through traditional seminars orthrough using web-based solutions (e-learning). Training can be classified in “trainthe user” (normally only on applications) and “train the administrator”. These aresimilar categories to those used in the support.

Differences between proprietary software and OSS

There is not inherent difference between both kinds of software. Training is done bytwo types of companies: Classical training companies offer training for all kinds ofsoftware including OSS products. For them, training is the core competence. Soft-ware distributors and service companies offer training for their software products andsolutions in particular. For them, training is one service that they offer, but normallynot their core competence. For OSS products, training is offered by distributors andintegrators.

Support

Definition of supportSupport is a form of enabling the users and administrators of software to get the soft-ware performing in the way it was supposed to perform (fixing problems). This canbe done through traditional telephone hotlines, e-mail, voice over IP or throughnewsgroups and message boards. There is user support and administrator support.


Support for proprietary software is offered by special service companies or by the soft-ware producer. Typically there is a distinction between private and corporate users.While private users are being offered the basic help functions within a program andsimple telephone support, corporate users can often obtain more extensive supporttargeted at administrators or even at helping with customisation.

Support for OSS products is first of all offered by the OSS community through var-ious forums. However, for corporate users this kind of support is not sufficient asthere is no guarantee for a problem being solved. Distributors and independent OSSservice and integration companies offer support. Some of the distributors employhome-working OSS developers who are experts for certain problems or parts of soft-ware. By their contracts, these OSS developers are obliged to give expert support tothe distributors‘ customers.

Application Management

Definition of application management?

Application management includes all activities to ensure that a software is performingin the way it was desired to perform. It includes permanent checking of the followingquestions:

❑ What are the functional and resulting technical requirements?❑ Is an update of the software to a newer version available and needed?❑ How is data transferred to the updated version?❑ What sort of backup is needed?❑ What sort of log file handling is needed?


There is basically no differenc between proprietary software and OSS. Applicationmanagement can be done either within the company (IT department) or through anexternal service provider. A feature of application management can be remote admin-istration, which is usually offered by distributors and independent OSS service andintegration companies.



2.2.6 Software product categories

Different ways for softwareclassification

Software can be classified in several ways. As chapter 2 provides an overview on soft-ware product related issues, we identified a technical or functional classification ofsoftware categories. The classification shown in figure 2-5 offers a better understand-ing of software product categories. The software products can either be standardisedor individual.

This classification according to technical categories is only one possibility, althoughthe most often used on Open Source software sites. However, the strategic marketanalysis in Chapter 3 will use a different segmentation according to user demands, asthe users are the primary determinants of demand. The segments are slightly differentfrom the technical classification.22

Figure 2–5Technical and functional

software productcategories

22. This distinction illustrates a problem many of the Open Source based business models had.Coming from the technological side, the companies focused very much on product characteris-tics and less on the customers they were planning to address.

Software

Operating Systems and OS Extensions

StandardApplications

EnterpriseSolutions B

erle

con

Rese

arch

200

2

OperatingSystems

OS Systems& Development

Tools

ProductivityApplications

SME BusinessApplications

Networking/Internet

Applications

Entertainment& MultimediaApplications


313 Open Source and the software market: market segmentation and hypotheses

3 Open Source and the software market: market segmentation and hypotheses

3.1 The software market

The software market within the IT market

The IT market can be differentiated into four different market segments: hardwareproducts, hardware maintenance services, software products and services, Internetand processing services.

Figure 3–1Structure of the IT market

Source: Hoch et al., 2000.

Hardware market divided into four product segments

According to an definition from EITO (2001), the hardware market is divided intofour product segments.

❑ Computer hardware (server systems, workstations, PCs, PC and workstationadd-ons, e. g. PC printers and other PC add-ons),

❑ End-user communications equipment (telephone sets, mobile telephone sets,other terminal equipment),

❑ Office equipment (Copiers, other office equipment), and❑ Datacom and network equipment.

Hardware maintenanceThe hardware maintenance services market is basically supplied by hardware produc-ers, by hardware vendors or by special service companies.

IT-Market

Ber

leco

n Re

sear

ch 2

002

EmbeddedSoftware

and Services

SoftwareProducts

ProfessionalSoftware-Related

Services

EnterpriseSolutions

HardwareProducts

SoftwareProducts

and Services

Internet and Processing

Services

HardwareMaintenance

Services

PackagedMass-Market

Software


3.2 Open Source and the software market: market segmentation and hypotheses32

Internet and processingservices

The Internet and processing services market is important for special service providers,xSPs (e.g., ASPs, payment service providers or B2B marketplaces) and similar busi-nesses. Their main business focus is different from software development. The differ-ent business focus leads to different income models. Most income comes from serviceor usage fees.

These three segments do not need to be analysed in detail. We will limit the analysison the segments and strategies, where OSS has a significant influence. It will becomeclear that between the four market segments several dependencies exist. The lines be-tween the market segments blurr since, for example, some players from the hardwarebusiness shift to the services and to the software business. We will put the main focuson the analysis of the software market.

The market segments

Three software marketsegments

As shown in figure 3-1, the software market can be divided into three segments:

❑ Software products (further to be separated in enterprise solutions and pack-aged mass market software),

❑ Professional software-related services, and❑ Embedded software and services.

The market segments will be analysed in detail in section 3.2. For each market seg-ment, we will firstly explain the business dynamics in detail, and secondly apply themto the identified market segments by characterising segments, identifying the majorplayers in the segments, and eventually generating hypotheses about the influence ofOSS in the segments.

3.2 Business dynamics and influence of Open Source soft-ware

3.2.1 The software products market

Customer-driven marketsegmentation

Section 2.2.6 provided a technological and product-driven segmentation of software.For the market analysis, we will adopt a customer-driven segmentation. We will seg-ment the market according to customer purchase decisions. For example, the marketfor operating systems on the desktop will not be regarded as a separate market becausethe purchase decision for an operating system for the desktop itself is usually notmade separately but always related to either the hardware available or the applicationsavailable.

Distinction betweenenterprise solutions and

packaged mass marketsoftware

As shown in figure 3-2, we will make two basic distinctions: The first one is betweenenterprise solutions and packaged mass market software. The market for enterprisesolutions is substantially different from the market for packaged software because therevenue portion based on service fees is considerably higher. Therefore, the marketfor enterprise solutions can be mainly characterised as service market. But the servicesales are primarily based on a special software developed.


333.2 Business dynamics and influence of Open Source software

Figure 3–2Structure of the software products market

Distinction between server and desktop market for packaged software

The second distinction is within the market for packaged software. We distinguishthe server and the desktop market. Within these segments, one could additionallysegment between operating systems and applications. But due to the fact that the pur-chase decision for the operating system is dependent on the decision for applicationsand vice versa, we will look at the server market in general and only when necessaryseparate between operating systems, applications or even appliances. Similarly, wewill look at the desktop market in general.

Figure 3–3Software for the server and for the desktop/client market

Overall three market segments

Together with the segment enterprise solutions, we have three market segments:

❑ Server operating systems and applications,❑ Desktop/client operating systems and applications, and❑ Enterprise solutions.

Business dynamics in the software products market

Analysis of the software business by McKinsey

In 1999, McKinsey analysed the software market and identified general business dy-namics in the product business (Hoch et al., 1999). The analysis was focused on thetraditional software business (development of proprietary software). The followingconclusions were derived:

Low entry barriersIn the software product business, entry barriers are low. The market is knowledge-driven, only low capital investment is needed. The low financial entry barriers boosthigh innovation rates (short time between releases), which, in turn, lower technicalentry barriers. Therefore, firms in this market face a constant threat of new entrants.

Software Products

EnterpriseSolutions

PackagedMass Market

Software Ber

leco

n Re

sear

ch 2

002

DesktopMarket

ServerMarket

Ber

leco

n Re

sear

ch 2

002

Server

ServerOS

Serverapplications

Desktopapplications

DesktopOS

Desktop

Serverappliances

-Operating

System andApplications

Applications

OperatingSystem (OS)



Low marginal costs buthigh fixed costs

Low marginal costs result in worldwide markets, high fixed costs for research and de-velopment result in the condition that many copies must sell. The latter statementbecomes blurred when considering OSS development. While the fixed costs are highin terms of hours devoted to a specific project, there are not necessarily monetarycosts involved. Thus, an Open Source software could – at least in principle – survivedespite a combination of large efforts and insufficient usage. Whether the program-mers find this appealing in the long-run and will not switch to other, more successfulprojects, is another question.

Tendency towardsconcentration

There is a race for leadership and a tendency towards concentration because of thefollowing aspects: Firstly, there exists a network effect because of the interoperabilityof programs. The ability of programs to operate and communicate with each other iscrucial. Hence, the more users a program has, the more the users can benefit from it.Secondly, there is a barrier for people to switch once they are trained to work with aprogram. Thirdly, popularity of a program is a major factor for the purchase decision(again a network effect).

The tendency for concentration seems to be stronger in consumer markets, wherepeople only execute programs. In markets, where the purchase decisions are made bypeople with special IT knowledge, the training barrier and the popularity barrier areusually not as high. People with IT knowledge base their purchase decisions to a lesserextent on popularity. They also get used to a new program more easily.

Unstable market position But any market position of a player in the software products market is never stablebecause of technology switches.

Server operating systems and applications

High-end… The market for server operating systems and their applications can be divided intodifferent segments. We have to look at the high-end server market with different play-ers, where the interdependence with hardware is inevitable because new hardware de-velopments (such as the Itanium processor) consequently lead to various projectsporting various operating systems to the new hardware.

… and low-end servermarket

The other segment is the low-end server market with a variety of applications. In thismarket, we have to look at the operating systems on one hand and on the server ap-plications on the other hand. The applications include:

❑ Intranet servers (database servers, file servers, print servers),❑ Multi usage servers (web servers, mail servers, streaming media servers, chat

servers), and❑ Connectivity servers (firewalls, gateways/routers, dialup servers).

Customers differ fromdesktop market

The customers in the server market are different from the customers in the desktopmarket. The decision makers have in general a basic IT knowledge and are corporatecustomers. Purchase decisions are made by management (small companies), IT man-agers (small and medium companies) or by IT departments (medium and large com-panies). The purchase decisions are made on criteria different from the desktopmarket. Therefore, the server market is driven by different critical success factors.

Most corporate customersbuy solutions

Most corporate customers buy solutions rather than mere products. Hence, servicecompanies, ISVs (independent software vendors) and VARs (value added resellers)are important distribution channels. Partnerships and co-operations with thesegroups are one critical success factor.

There is a discussion about the criteria that determine the purchase decision in theserver market. Basically, they relate to the technical and functional quality of the soft-ware, but there are also other aspects like the number of skilled developers available.



Products and players in market for server operating systems

Important products and players in the high-end server operating systems market are:IBM (AIX, AS400, OS/390, OS/400), Fujitsu-Siemens (BS2000, BS2000/OSD),Sun (Solaris), HP (different Unix versions), Silicon Graphics (Irix), Compaq/Digital(Unix, VMS). The main products and players in the low-end server operating systemsmarket are Microsoft (Windows plus server applications) and Unix in different vari-ations and derivatives (one of which is Linux plus applications).

OSS influences on the server software market

Linux is supposed to offer several advantages and to constitute a serious alternative asserver operating system due to the following advantages.

❑ High stability and high reliability,❑ Low/zero license costs,❑ Freely modifiable features, and❑ Relatively small kernel.

Unix applications relatively easy to transfer

Furthermore, many applications for Unix systems are relatively easily transferable toLinux. Also many software application vendors are interested in having their Unix ap-plications to run on Linux and vice versa as support of a single product is muchcheaper than supporting different versions.

Thus, there are a number of factors and developments that support the hypothesisthat Linux market share in the server market is going to rise further.

No restriction to LinuxIn the server applications market, we are not only talking about Linux. It is the webserver software Apache that leads the market of web servers on the public Internetwith a market share in September 2001 of about 61 percent. Microsoft had a marketshare of 29 percent and iPlanet (Netscape) 2 percent (Netcraft, 2001). Besides theweb server combination of Linux and Apache also the file server combination Linuxand Sambahas significantmarket share.

Desktop/client operating systems and applications

Classical mass marketThe market for desktop/client operating systems and their applications is the classicalmass market. It is also the market that is mostly referred to when talking about soft-ware markets. The market is a worldwide market. It has a high tendency towards con-centration and is dominated by Microsoft. Customers base their purchase decision onpopularity, on interoperability with other systems and applications, and on numberand quality of applications available for the operating system. In addition, they faceswitching costs due to training. Therefore, the market leader has a strong position.

Service companies and VARs (value added resellers) are an important sales channelfor private and SME sales because many customers buy computer equipment with theoperating system and the applications pre-installed. Hardware companies and theirsales channels are important because of the integration of software and hardware (pre-installation) and the bundled sales. For software producers, the partnerships and co-operations with hardware producers and with service companies and VARs are criticalsuccess factor.

Major actors on the marketThe main players and products with regard to desktop/client operating systems are(market shares according to IDC, 2000a):

❑ Microsoft (Windows) dominates with a 88 percent market share in 1999.❑ Apple (Mac OS, Mac OS X) comes second with a significantly lower market

share of 5 percent in 1999.❑ Linux on the desktop has a market share of 4 percent in 1999.

Since the time this data was generated, the market share of Microsoft has even in-creased to well above 90 percent.



Segmentation ofapplication markets

The applications in the desktop/client market can be segmented along several criteria.The most simple differentiation is into horizontal (the general office programs) andvertical (functionally specific or industry-specific) applications. As already men-tioned, the market for applications is closely linked to the market for operating sys-tems. Therefore, the players are basically the same. The market leadership forapplications is held by Microsoft with its Office Suite, which also runs on many Mac-OS desktops. There are various OSS applications, which run on Linux or other OSSoperating systems, but also on Windows.

OSS influence on thedesktop/client software

market

There are different possibilities of competition between the market leader Microsoftwith its Windows and Office bundle and Open Source software. First of all, there isdirect competition of bundles. For a long time Linux in combination with OpenSource programs (e.g., OpenOffice) has been positioned as competitor. As these ap-plications do, however, not provide the same level of user-friendliness and familiarityas the Microsoft family, strong competition is unlikely. The unequal positions areeven strengthened by the fact that Microsoft products are sold pre-installed with mostIntel-based computers and many (private and SME) users buy a bundle of hardware,operating system and applications.

This has changed with MacOS X which, although not purely Open Source, is based(in parts) on an Open Source Unix. Together with the available Microsoft Officesuite, this product bundle of hardware, operating system and applications provides astronger competition to the Wintel bundle. Although not purely Open Source, theadvantages of the Open Source development model might make the Apple alternativemore powerful and might increase Apple’s chances to compete more strongly withMicrosoft. The outcome of this stronger competition has to be awaited.

A further influence of Open Source can be in the field of desktop applications. Severallarger desktop application projects, e.g. Mozilla or OpenOffice, have reached a stage,where it is expected that they can become strong competitors to the respective Micro-soft products Internet Explorer and Office within one or two years.

Enterprise solutions

ERP systems and otherenterprise applications

The major products in the enterprise solutions segment are ERP (enterprise resourceplanning) systems. But several other products belong to this category as well, e. g.,CRM (customer relationship management) software, SCM (supply chain manage-ment) software, KM (knowledge management) software, groupware, e-learning soft-ware, etc. Worldwide, the main players are according to Hoch et al. (1999): IBM,Oracle, Computer Associates, SAP, HP, Fujitsu-Siemens, Hitachi, Parametric Tech-nology, Peoplesoft, and Baan.

Customisation needed Enterprise solutions almost always need customisation. Hence, the market is charac-terised by revenues that are based on product licences on one hand and service feeson the other hand. According to an internal McKinsey study, 30 percent of the in-stallation costs for an ERP system account for the software product licence, 70 per-cent account for professional services fees to implement the product. (Hoch et al.,1999, 36). Some software companies offer the service themselves, some through part-nerships. Customisation and installation projects typically take several months to becompleted.

As this market is somehow between the products and the services market, we will in-clude the market for enterprise solutions in the analysis of the services market.



3.2.2 The market for software-related services

Business dynamics in the market for software-related services

As (Hoch et al., 1999, 42) have found out in their analysis, the business dynamics inthe market for software-related services differ in several ways from those for the soft-ware products market, mainly due to its service character.

Low entry barriers, constant threat of new entrants

The capital to be invested to start a service company is low. Instead, knowledge is themajor foundation and entry barrier. Therefore, it is easy for a new company to set upif its founders have the technical knowledge to offer software services. Thus entry bar-riers are low, and incumbents constantly have to be aware of new entrants.

High pace of innovationA high pace of innovation in the software products market leads to a similarly highpace in the services business. New technologies arise and with them new companies,which are able to offer the service. Some players in the services business are offeringproducts as well.

Constant and significant marginal costs

Software services are a classical people-selling business with constant and significantmarginal costs. In this sense, the services market is different from the product market.The cost of a second project are quite the same as the cost for the first project, evenwhen a similar solution is implemented.

Human resources Human resources are the most important asset in the services market. Therefore, it isimportant to attract capable human capital and to invest in human resources.

High fragmentationContrary to the product business, the services market is not ruled by the law of in-creasing returns. Therefore, the market faces much higher fragmentation and one canfind:

❑ Small companies,❑ Regionally focused companies, and❑ Very few truly global players in service firms (e.g., Accenture, IBM Global

Services, EDS, CSC, Cap Gemini Ernst & Young, KPMG, Pricewaterhouse-Coopers, Deloitte & Touche, Cambridge Technology Partners).

Market segments for software-related services

Two basic market segmentsThe services market consists of two basic market segments:

❑ IT consulting and systems integration as well as❑ IT services.

The firms in the IT consulting and systems integration segment provide mainly cus-tomised services (solutions), including customised software products.

The IT services include support and maintenance, training and application manage-ment (including outsourcing). The firms in this field provide mainly these services,but they can sometimes also offer consulting and systems integration. The IT servicesseem to be stronger related to a product business. Many of the larger firms in this fieldare service units of hardware producers, sometimes even separate entities, such as Sie-mens Business Services or IBM Global Services. Others are primarily software pro-ducers where the focus on services becomes more important and generates a higherportion of revenues than the software sales.

There is no clear line between the two market segments, and many companies are ac-tive in both segments. The two segments are represented in the business structure ofsome firms in the service market. They usually have a business unit for “corporate so-lutions” (which includes consulting, implementation, and integration). The business



is project-related – normally a project is finished once a new solution is implementedand the customer is trained to work with it.

Major companies Major firms active in these services markets are:

❑ The global service companies that usually offer services to large (global) enter-prises.❑ Worldwide: Accenture, IBM Global Services, EDS, CSC, Science Appli-

cations, Cap Gemini Ernst & Young, Hewlett Packard, DEC, BSO Ori-gin, Fujitsu-Siemens (Hoch et al., 1999, 27).

❑ Germany: CSC Ploenzke, CAP Gemini Ernst&Young, Accenture, gedas,PricewaterhouseCoopers Unternehmensberatung, KPMG Consulting,Plaut Gruppe, Atos Origin, IBM Deutschland, Siemens Business Services,T-Systems, Hewlett-Packard, Lufthansa Systems Group, GE CompuNetComputer (Lünendonk, 2000)

❑ Thousands of small service firms, either focused on specific solutions andtechnology or on their home region. The latters’ main customers are oftenSMEs.

OSS influence on theservices market

With regard to the global services companies and their large integration projects –such as enterprise solutions – OSS will not yet have a significant influence. The OSS-related discussion is mainly focused on the question whether the main firms in themarket of enterprise solutions will make their solutions portable on Linux. (SAP,IBM, Oracle, Software AG, Sybase all have already done so with major ERP applica-tions.)

With regard to small service companies, however, OSS has a significant influence.Firstly, there are service companies that focus exclusively on Linux and other OSS.Secondly, there are distributors of OSS that offer services and support for their prod-ucts. Thirdly, there are specialised service companies that shift to OSS or includeLinux and other OSS into their spectrum. Strengths and weaknesses as well as thethreats and opportunities of the three groups in relation to each other will be analysedin section 4.2.3.

3.2.3 The market for embedded software products and services

Business dynamics in the embedded software market

Customers are producers ofspecial devices

The major difference between the software products and embedded software is thecustomer. In the market for embedded software, the buyers of software are not theend-users, but the producers of different kinds of devices. Hence, the direct custom-ers of embedded software are software developers who need to work with embeddedsoftware products and develop their customised version of software to ensure func-tionality in any kind of device. Therefore, the market for embedded software prod-ucts is a classical B2B market. The software companies that supply embeddedsoftware are the suppliers for the device producers.

Major business dynamics We will limit the analysis to embedded products and, in this first step, to the embed-ded operating systems in particular. The business dynamics in the embedded softwareproducts market are similar to the dynamics in the non-embedded market. The majorbusiness dynamics can be summarised as follows:



❑ The indirect network effect is strong because of the interoperability and port-ability of software and because of the skills of developers. The less commonany embedded system tool is, the fewer common components can be used andthe fewer people are able to handle the embedded software. The network ef-fect is especially strong for operating systems, which form a basis for manyother software applications and software development tools.

❑ The business is knowledge-driven and the capital investment is low, which re-sults in high innovation rates. However, the network effect establishes a highentry barrier.

❑ The marginal costs are low and result in worldwide markets. With non-OSS,the fixed costs for development are high and result in the condition that manycopies must sell in order to have a profitable business.

❑ In the long term, the market position is never stable because of technologyswitches.

Analysis of the embedded software market

Home-grown operation systems with traditionally large market share

Traditionally, the largest share of the embedded operating systems and kernel marketis occupied by so-called “home-grown” operating systems. These are operating sys-tems, which are privately developed and maintained. They used to account for be-tween one half and two thirds of all embedded systems’ operating systems (EvansData Corporation, 2001). According to a survey among 500 developers in 2001,home grown systems are still the most widely used systems, followed by WindRiver’soperating systems and DOS. Embedded Linux, as an open operating system, is al-ready number four.

Challenges for market leaders

The home-grown systems face several difficulties, which will become even stronger inthe near future. Basically, applications in the embedded market are getting muchmore complicated, and networking is becoming more important, which makes devel-oping and maintaining these private systems much more expensive. Therefore, it ispredicted that many of the home-grown operating systems will switch to Linux in thenear future. (Evans Data Corporation, 2001) Half of the home-grown systems oper-ate real-time, half of them are smaller, non real-time systems. A real-time operatingsystems needs to guarantee that a given operation will be performed within a giventime window. Hence, the critical factor for real-time is determinacy.

Advantages for Linux …Linux is supposed to have several advantages compared to the home-grown systems(Cook, 2000):

❑ It is standardised and Open Source. It has a modular structure and can betrimmed down for several purposes. Hence, the features of Linux are freelymodifiable for various reasons.

❑ It has a low price.❑ There are many skilled programmers available on the labour market.❑ Compared to private operating systems, there are many drivers available.

(Compared to Windows, of course, not so many.)… and disadvantagesHowever, there are also several disadvantages:

❑ Linux is not designed to be a real-time system. Although there are variousprojects to develop components to make Linux real-time for various situa-tions, the results are not comparable to operating systems that have been de-signed for real-time performance.

❑ Compared to other proprietary embedded operating systems, there is still alack of drivers for Linux.

❑ Linux is relatively large by embedded standards. However, it can be trimmed



down and there is a trend for hardware products becoming cheaper. And thekernels of its major competitors offering proprietary software are also quitelarge.

❑ Embedded systems are often mission-critical for business processes. Here, theperceived lack of professional 24h support is a disadvantage for Linux.

OSS influence on theservices market

Linux constitutes an alternative for home-grown systems when their operations notnecessarily have to be real-time and when their operations are not mission-critical. Anumber of Linux embedded products is expected, such as set-top boxes, cable TV,toys, car devices, manufacturing devices, in the long-term also house-hold devices.(IDC, 2000a) Here, Linux has potential to gain market share.

The major suppliers of embedded Linux are seen to be Lineo, MontaVista and RedHat. (VDC, 2001)


414 Business models based on Open Source software

4 Business models based on Open Source software

4.1 Overview

In our analysis, we identified several types of business models that are based on theOSS technology. Of course, in reality, many companies operate in more than onefield, for example the distributors. Although their main business is marketing and dis-tribution of products, a major portion of their income is based on additional servicessuch as consulting and support.

Focus on business models that are purely based on OSS

The focus of the subsequent analysis is on business models that are purely based onOSS. This means they would not exist without the occurrence of the OSS phenom-enon. Similar to the software market analysis, we distinguish between product-relatedbusinesses and service-related businesses. Figure 4-1 provides an overview of the ex-isting business models. Service and support providers offer consulting, system imple-mentation and integration, support, training, recruiting and staffing services.

Figure 4–1OSS business models

Questions for analysisThe business models will be analysed in the following section with regard to the fol-lowing questions:

❑ What do the companies do (main product or service offer)?❑ What is the market (customer and competitive analysis)?❑ Why do they earn money (Why not)?

OSS Business Models

Berlecon Research 2002

Distributors and Retailers

OSS-relatedServices

Niche andSpeciality OSS

Distributors

Original LinuxDistributors

Retailers of OSSDistributions andCompl.Products

OSS Developmentand Interest

Enablers

Service andSupport

Providers


4.2 Business models based on Open Source software42

4.2 Distributors and retailers

4.2.1 Original Linux distributors

Product and service offerings

Examples The Linux distributors package and sell their own version of the Linux operating sys-tem. Examples for original Linux distributors are Red Hat, SuSE, MandrakeSoft,Caldera, Turbolinux, and Slackware.

To end users (consumer or corporate), they sell the Linux operating system in varioussoftware packages and bundles such as server applications, desktop applications, e-commerce suites, often in variations for various hardware. To IT administrators, theysell appropriate administration tools for their applications. To developers, they selldevelopment tools and usually also various Linux versions for OEMs (original equip-ment manufacturers), which use the operating system as a basis for their hardware-software combination packages or embedded systems.

Elements in Linuxdistributions

A Linux distribution consists of the Linux kernel and several hundred additional filesthat together form the Linux operating system. In order to develop their own ver-sions, the distributors need to collect the newest Linux release and all the related files.The second step is testing, tuning and optimising the existing software fragmentsworking together with the aim to achieve a good performance and reliability. Theseefforts are normally returned to the OSS community. In a third effort, means forsmooth installation, good documentation, efficient management and productivitytools are created. For these steps, the Linux distributors support the Linux commu-nity by providing development laboratories. Additionally, they employ several devel-opers or work with freelance developers.

Pros and cons of beingdistributor

On one hand, Linux distributors save an enormous amount of software developmentcosts because they do not have to develop their operating system from scratch. Forexample, the Red Hat 7.1 operating system is estimated to have cost $1 billion(Wheeler, 2001a). However, there remains a significant investment for the develop-ment of their optimised Linux versions. On the other hand, the Linux distributorscannot price their products as freely as proprietary software producers can. The sev-eral components of Linux can be downloaded freely and as many times as possiblefrom several project websites. Therefore, the value added by distributors on the prod-uct level is mainly the packaging.

The software products based on the Linux version are eventually offered on a websitefor download or on a CD-ROM for physical distribution. The Linux distributors useseveral sales channels – the most important ones are the VARs (value added resellers)and retail chains (in particular bookstores).

Critical success factor A critical success factor in the Linux distribution business is brand building. Hence,the distributors heavily invest in marketing (advertising, trade fairs, public relations).Distribution and Marketing are the core competences of the Linux distributors.

Additional services Despite this fact, most of the distributors provide additionally Linux-related servicessuch as consulting, integration, support and training. The services generate an addi-tional income stream. Additionally, the distributors generate a small income streamthrough merchandising – the selling of T-shirts, mugs etc.



The market for Linux distributors

Two market segments covered by distributors

Basically all of the Linux distributors cover two market segments: First, the mass mar-ket with standardised packages offered to SMEs and private consumers. Second, themarket for individual solutions, which are offered to medium to large corporate cus-tomers.

The mass market for operating systems separates into the server and the desktop mar-ket. In the server market, OSS is supposed to offer several advantages and to consti-tute a serious alternative as server operating system. The major competitors areWindows NT and the various other Unix systems. In the desktop market, the Linuxmarket share is very small. Here, Linux has one major competitor: Microsoft with itsvarious Windows versions. The question for the coming years is whether the Linuxoperating system will be successful on the desktop or not.

The solutions market (service-related) is completely different from the mass market(product-related). Most of the distributors consider the solution business as increas-ing and profitable income stream. Some of them cover the solutions market throughpartnerships with consulting companies. On the other hand, companies like SuSEhave build up their own profit centre for corporate users.

Advantages and disadvantages of business model

Low margins for distributors

In the mass market, Linux itself can be regarded as a commodity because its compo-nents can be freely downloaded or copied. The packaging, which is done by every dis-tributor and results in the different Linux versions, constitutes the added value.However, the margins per unit sold are not very high. The product business withLinux distributions is a mass market and the active companies will have to increasetheir product sales (and eventually their market share) to become profitable.

Differentiation by branding…

Therefore, the distributors are forced to develop other means to differentiate them-selves from each other and from proprietary competitors. So far, this differentiationhas been mainly achieved by branding, a critical success factor in the mass market.Linux distributors such as Red Hat or SuSE put much effort into marketing (adver-tising, trade fairs, even certification of their Linux trainings can be regarded as a brandbuilding effort).

…and exclusive sales channels

The second success factor is gaining access to sales channels, which are bookstores andVARs. Mandrakesoft, for example, exclusively cooperates with Macmillan bookstoresin the US. In Germany, the company is trying to intensify partnerships with smallconsulting companies and integrators to gain access to the SME segment.

Distributors shifting towards solutions and consulting

As it is difficult to survive on the product business alone, the distributors build up asecond business by shifting towards the solutions and consulting market. This mightbe interpreted as a move towards a more lucrative business as pure software retailingprovides only low margins because the marketing costs are high and, in the desktopmarket, the number of potential buyers of Linux box products is still low comparedto the buyers of Microsoft box products. Due to their software knowledge from pack-aging and optimising the Linux parts, the distributors certainly have the OSS as wellas technical competence to build up consulting and service business. But it is in ques-tion whether they already have the know-how in consulting and business processes tobecome serious competitors for existing service or consulting firms.



Figure 4–2Primary focus of

distributors in the softwarevalue chain

4.2.2 Niche and specialty OSS distributors


Specialised on OSS otherthan operating systems

The niche or specialty distributors develop and distribute different OSS but no oper-ating systems. Their products include applications, development and administrativetools. Normally, their software is developed to run on Linux, but some products alsorun on Windows or other operating systems. Examples are Zope (formerly DigitalCreations), Sendmail.com, Covalent Technologies, Cygnus (acquired by Red Hat),Precision Insight (acquired by VA Linux), MySQL, ActiveState, and CollabNet.

Companies livesymbiotically off an OSS

project

In this business model, companies live symbiotically off an OSS project. OSS is col-lected, maintained and/or developed. The main function of those companies is to co-ordinate the scheduling and make a commitment to delivery and support of adedicated product. Normally, they employ some of the core developers of the specificproject and rely heavily on their relations to the developer community.

Some examples in detail: Zope offers an OSS applications server, a platform to devel-op CMS (content management systems) among others, and several additional toolsto add task-specific or industry-specific features. ActiveState offers proprietary devel-opment tools for Linux development, CollabNet proprietary software developmentmanagement tools. Precision Insight offers OSS server tools to support graphics hard-ware. Covalent Technologies offers an optimised version of the Apache web server.Sendmail.com offers a message server and a version of its software to be embedded invarious products. Similarly, MySQL offers its database software in a version for com-mercial use and to be embedded in other software products.

The market for niche and speciality OSS distributors

Market share often high The market share for some OSS products such as Samba or the Apache web server israther high. For example, a web server survey conducted by Security Space countingthe web servers across all domains, reports an Apache market share of 63 percent inOctober 2001. Similarly, a web server survey conducted by Netcraft counting activeserver across all domains reports an Apache market share of 61 percent in October2001. However, this does not say much about the success of the companies that aretrying to profit from this development.

Customers often VARs orOEMs

The market access for the niche and specialty distributors is very different from theaccess for the Linux distributors. Normally, they do not directly target private con-sumers or SMEs with their products. The major Linux distributors can easily incor-porate specialty software components into their out-of-the-box packages and, becauseof their stronger brands, sell it to a broad customer base. Nevertheless, some specialtydistributors offer a limited number of packages that can be ordered on their websitesand are distributed through the same channels as the Linux distributions (e.g. VARsand retail chains).

Software Development

Software Packaging

Ber

leco

n Re

sear

ch 2

002

Marketing and Sales


tation/ Integration


Management

Software Documen-

tation



Hence, the major customers of specialty distributors must be either VARs (ValueAdded Resellers) or OEMs (Original Equipment Manufacturers), which sell opti-mised hardware-software bundles or develop and sell embedded products.


Because the companies in this category develop and distribute OSS, which can nor-mally just be downloaded and copied, it is hard to imagine what kind of direct busi-ness model they could use. Indeed, simply selling software does not propose a uniqueselling proposition because access to OSS is largely unrestricted, at least in most li-cense models.

Variety of income streams based on complimentary products

As a result, there are a variety of income streams summarised in this model. Most ofthe companies sell additional services for their product (consulting and support).Some of them, for example MySQL, generate income from license fees for commer-cial licenses for the normally GPLed MySQL. Others, for example Precision Insight,decide not to disclose the source code for the newest version of their products, butonly for previous versions. Another way, used by Sendmail.com, is to develop propri-etary commercial software on top of the basic sendmail functionality. Most of theseideas are critical because the companies leave the field of pure OSS players and be-come a player in the traditional software business. They have to be able to live in bothworlds.

4.2.3 Retailers of OSS distributions and complementary products


Sales channels for software and documentation

The retailers are major sales channels for the distributors. They either sell the distrib-utors’ software products or they provide and sell additional documentation and in-formation on OSS products or merchandise. The retailers are not involved in thesoftware development process. Instead, their core competence is distribution or, insome cases, publishing. The retailers are not solely focused on OSS. For LehmannsFachbuchhandlung or mitp, for example, OSS training and documentation books arejust one part of their retailing or publishing business. O’Reilly is the only knowncompany, whose business is almost solely based on OSS. O’Reilly is selling documen-tation and training books.

Other examples for publishers and retail chains include CNET, ZDNet, and Fona(Denmark). In addition, there are several specialised Linux shops like CrazyPenguin(UK), Linuxland (Netherlands), Linux Central (U.S., part of INT Media Group),and Linuxbutikken (Sweden, Norway).

The market for retailers and publishers

Firms target the mass market

The retailers and specialised Linux shops target the mass market only. Their custom-ers are private or corporate users, developers or IT-administrators. For OSS, the mar-ket is slowly shifting to users that are not “software freaks” or developers, but insteaduse the software.




Access to customers andknown brand

The advantage of established retailers is their access to customers via their retail stores.Another advantage is the known brand of retail chains. Web-based resellers and spe-cialised Linux shops will find it difficult and expensive to create a brand awareness,especially when they compete directly with the distributors and publishers.

Merchandising, in general, is not a business model itself but only an additional in-come stream whenever a strong brand has been established. Therefore, income frommerchandising is primarily interesting for the Linux distributors.

Need for documentation There was a need for documentation as OSS became more popular and was appliedoutside the OSS community. For commercially developed software, the documenta-tion is normally done by the software producer or in cooperation with a publisher.O’Reilly combined its OSS knowledge with publishing knowledge and succeeded inestablishing a brand for OSS books. Because the company covers many OSS projects,it is not dependent on a single software development.

4.3 OSS-related services

4.3.1 OSS development and community enablers


This category includes primarily two different sorts of actors. These are first of allmarketplaces like SourceXchange, Cosource.com, intraDAT (vshop.org) and sec-ondly conference and trade fair organisers like LogOn Technology Transfer or LinuxNew Media.

Marketplaces The function of exchanges or marketplaces is to match potential buyers (organisa-tions or individuals looking for “needed improvement”) and sellers (OSS developercommunity). The software produced would be customised or build-to-order OSS.The main argument for the potential of these exchanges is the assumption that manysoftware developers want to decide themselves what project to work on (which is notpossible as an employee). Additionally, the global reach of the Internet could leveragethe developer potential all around the world and possibly even drive the prices down.

Improving developmentprocess with additional

services

The marketplaces for software development offer the matching service and improvethe development process through provision of a project manager and productivitytools. Even demand aggregation can be a service: Multiple buyers with the same prob-lem aggregate their funds via the marketplace to get a software solution. As far asknown, no company in this business has become profitable so far. SourceXchangeclosed in April 2001.

Conference organisers – the other business model in this group – enhance interest inOSS projects and provide opportunities for the OSS community and business part-ners to meet. The organisers either are specialised in OSS and Linux or are generalconference organisers that generate part of their income through their focus on OSS.There is some relation to the publishing business and consequently publishers likeO’Reilly are also active in organising Open Source conferences.



The market for OSS marketplaces and conference organisers

Chicken-egg problem for marketplaces

Marketplaces address two groups, the buyers and the suppliers of software. The “sup-plier” group is the OSS developer community, the “buyer” group are corporate usersor developers, who look for a specific solution. Normally, the buyers would be willingto pay for a service like those provided by the marketplaces. For the buyers, however,any marketplace is only interesting if there is a sufficient number of interested sellers,e.g. developers. So OSS marketplaces are facing a chicken-egg-problem like manymarketplaces in other areas.

Potential lack of sponsors for conference organisers

Conference and trade fair organisers address mainly the OSS community. This canbecome a problem because the business model does not really conform to the OSSphilosophy. Organisers in the traditional conference business address a group of buy-ers and sellers who are interested in product or technology information and know-how. They generate most income from the sell-side because sellers use the events formarketing their products. Additionally, buyers or people interested in gaining infor-mation pay entrance fees for conferences (high fees) or trade fairs (lower fees). It isquestionable whether the market for conference organisers in the OSS business is at-tractive.


MarketplacesSo far, the pure marketplaces and exchange models have failed. Probably the valueadded is not sufficient to carry a business model purely on the matching function.Revenues can only be generated from the demand side as the developers are probablynot willing to pay for the service.

On the demand side the “buyers” of the software might not have trust in the comple-tion of the projects. A normal software development project is usually done with asubcontractor according to a contract with exact specifications on what needs to bedone. These projects often exceed the cost and time in the contracts. With regard tothese experiences, it seems unlikely that a company would trust a developer commu-nity with no or vague responsibilities and certainly no guarantees for completion ofprojects. On the supply side, the main competition of these business models is theOSS community itself and all the projects that are managed by volunteers.

Potentially useful as complimentary service

The matching function could rather be used as an additional service in the spectrumof a service company. Some distributors, for example MandrakeSoft, apply part ofthis model in their support offerings. They pay OSS developers and specialists a cer-tain amount of money to solve specific problems.

Conference organisersSo far, OSS conferences and fairs have usually had comparatively low prices and it isquestionable whether the organisers can operate profitably. They cannot demandhigh entrance fees for conferences because most people interested are OSS commu-nity members. Nor can they demand high fees from software suppliers because manysuppliers are community projects or small and regionally focused service companies.At least, due to constantly rising interest in OSS, the trade fairs can become a moreattractive field because of rising visitor numbers.

Conference organisers are very dependent on the interest in a special subject. Theircore competence, however, is expert knowledge for the evaluation of interesting top-ics, finding interested people, and the project management of a conference or tradefair. With regard to those competencies, however, conference organisers do not needto restrict their business model to OSS. Although their knowledge can be OSS-lim-ited – in order to evaluate OSS topics, they need to have knowledge about the soft-ware market in general.



Reputation as entry barrier An entry barrier for potential entrants in the OSS conference market can be reputa-tion within the OSS community. Some OSS conference organisations have alreadygained significant reputation in the community. However, as mentioned before, thecritical factor in the OSS conference business is that a significant part of interestedvisitors is not willing to pay the corporate entrance fees that are paid on most B2Bevents.

4.3.2 OSS-related services and support

Whole variety of servicesoffered by companies with

different backgrounds

OSS-related services and support include several services such as consulting, systemsintegration, support, maintenance, remote administration, training, and applicationmanagement. The companies in the OSS-related services market differ according totheir background. First of all, there are companies that have a background in Linuxor other OSS products. They are trying to establish services that build on their prod-uct knowledge. Hence, their core competence is the technological and productknowledge. Most of them offer a full range of services. Linux distributors, niche andspecialty distributors as well as independent OSS service companies belong to thiscategory.

And secondly, there are companies that have special process knowledge in how toprovide a service related to IT in general. This can be knowledge in IT consulting,systems integration, IT-training or IT-recruiting, sometimes even with a verticalfunctional or industry-specific specialisation. They can extend their offerings to OSS-related services.

Examples Examples for full service companies offering various services based on OSS knowl-edge are:

❑ Linux distributors (e.g., Red Hat, SuSE, Caldera, MandrakeSoft, Tur-bolinux).

❑ Niche and specialty distributors (e.g., Zope, MySQL, Sendmail.com, Cova-lent Technologies),

❑ Independent OSS service companies (e.g., Linuxcare) and many small serviceand integration companies with special technological OSS and particularLinux knowledge (e.g., Linux Information Systems, B-connected).

Examples for special services based on integration and service knowledge extended toOSS are:

❑ For consulting and systems integration: (Global System Integrators; Accen-ture, KPMG, PricewaterhouseCoopers etc.), various small consulting and in-tegration companies.

❑ For training: various training and e-learning companies (e.g., Microconsult)❑ Recruiting and Staffing services: various IT-specialised recruiting companies

(StepStone-IT, Monster.de, JobUniverse.de)Figure 4–3

Primary focus of servicecompanies in the software

value chainSoftware

Development

Software Documen-

tation

Software Packaging

Marketing and Sales


tation/ Integration


Management

Ber

leco

n Re

sear

ch 2

002




Critical factor often Linux experience

Consulting companies and system integrators help their customers realise IT strate-gies based on business needs as described in section 2.2. The critical factor is oftenLinux expertise: The small integrators and service companies do usually have a back-ground in OSS development and are trying to establish a business based on services.The large integrators can activate their Unix people to acquire Linux expertise or hireOSS developers.

Support companies offer their services in various models, as the “classical” supportmodel of OSS (send a bug report to the project’s community) is not accepted bymany business customers. Commercial support gives businesses the possibility tohave their OSS products supported without getting involved with the developer com-munity culture (even if the support staff are developers, e.g. at MandrakeSoft). Asshown by IDA/Unisys (2001), support contracts vary in terms of...

Variation of support contracts

❑ Contract categories: installation support (for a limited time after installation),support packages (price per call), annual support contracts.

❑ Way of support: telephone hotline, e-mail hotline.❑ Level of support: 1st level (smaller problems, end-user targeted), 2nd level

(administrator), 3rd level (developer targeted, sometimes including sourcecode changes) support definition and coverage.

❑ Coverage hours and days: 10x5 to 24x5 or 24x7, working days only or all 365days/year.

❑ Reaction time: from 1 hour in the best case to 8 or 16 hours (or “next businessday”).

❑ List of supported products: hardware and/or software.❑ Personalisation – inclusion of individual consultancy and auditing.❑ Patch- and update management (in case of new version).❑ Supported infrastructure type, from desktop-PC to mainframe.

OSS focused training companies have two choices offering their courses: They can of-fer classical seminars with physical attendance (often after a certification process bythe software vendor) or e-learning solutions (e.g. Red Hat).

The market for OSS-related services and support

Customers for systems integration,…

The customers for systems integration range from small to large corporations, whichpay for a solution instead of paying for a product. Hence, the service is project-relat-ed.

…support, …Support is needed in any market and on any user level. For example, OEMs (originalequipment manufacturers) and ISVs (independent software vendors) can be custom-ers of Sendmail.com support when incorporating Sendmail into their product. Sys-tem administrators do usually need support when a new product is implemented. Butalso private and business users need support with their product (which is normallyoffered in a standardised way).

…and trainingCustomers for OSS training are users on various levels: Red Hat, for example, is of-fering courses for users, systems administrators and developers in classical seminars aswell as in e-learning courses. Customers typically are business-related users. Trainingproducts at Red Hat have one focus on the Red Hat Linux distribution and relatedsoftware; under “E-Business” they offer a course for SAP-Red Hat integration. (Intheir e-learning courses, they also offer C/C++/Java programming and general Unix/networking courses.)




Success depends onimportance of productknow-how vs. process

know-how

As already mentioned, there are two fundamentally different groups of firms active inthe OSS-related services market. Firms with OSS background have substantial prod-uct and technology knowledge, which they use to build up their services business.Businesses solely based on OSS products are dependent on the acceptance and devel-opment of OSS. Companies without OSS background have substantial processknow-how in the services sector. They attempt to extend their offering to OSS-relatedservices. Whether firms from one group or from the other group will succeed, de-pends on the importance of product know-how vs. process know-how in the separateservice fields. Figure 4-4 illustrates this relation.

Figure 4–4Areas of success in the

OSS-related servicebusiness

Strategic consulting vs.product support

Strategic consulting is a service field where methodology and process know-how areextremely important, product know-how is less important or, for a special project,can easily be acquired. Product support, on the other hand, requires primarily prod-uct know-how whereas knowledge about the support process can as easily be acquiredas process know-how in the consulting business.

The firms with OSS background will be mainly successful in areas where productknow-how is important and process know-how can be easily acquired. This is the casewith support and training offerings. The players without special OSS know-how willbe mainly successful in areas where this know-how plays only a minor role or can eas-ily be acquired.

Ber

leco

n Re

sear

ch 2

002

StrategicConsulting

ITConsulting

Trainingand

Recruiting

ProductSupport

Solutions market Mass market

Product andSoftwareKnow-howimportant

Process andMethodologyKnow-howimportant

Impo

rtan

ce o

f Kno

w-h

ow

Focused service companies

Full-service distributorss andOS/FS-related service companies

Number of Products sold


515 References

5 References

5.1 Literature

Balzert, Helmut (1996): Lehrbuch der Software-Technik: Software-Entwicklung,Heidelberg.

Balzert, Helmut (1998): Lehrbuch der Software-Technik: Software-Management,Software-Qualitätssicherung, Unternehmensmodellierung, Heidelberg.

Bezroukov, Nikolai (1999a): A second look at the Cathedral and the Bazaar, in: FirstMonday, Vol. 4, No. 12, 1999, http://www.firstmonday.dk/issues/issue4_12/bezroukov/index.html.

Bezroukov, Nikolai (1999b): OSS Development as a special type of academic re-search (a critique of vulgar Raymondism), in: First Monday, Vol. 4, No. 10,1999, http://www.firstmonday.dk/issues/issue4_10/bezroukov/index.html.

Blase, Paul (2000): Open Source White Paper – A Software Company’s Dilemma.DiamondCluster, http://www.diamondcluster.com/work/Wpapers/WPSoft-ware.asp.

BMWi (2001): Open-Source-Software – Ein Leitfaden für kleinere und mittlere Un-ternehmen, http://www.bmwi.de/Homepage/download/infogesellschaft/Open-Source-Software.pdf.

Claybrook, Bill (2001): AberdeenGroup InSight: Linux is on the Move Up!. Aber-deen Research, July 24, 2001, http://www.ibm.com/linux/LinuxInSight.pdf.

Cook, Rick (2000): Embedded Linux Basics. LinuxWorld.com, May 2000, http://www.linuxworld.com/linuxworld/lw-2000-05/lw-05-embedded.html.

Dafermos, George N. (2001): Management and Virtual Decentralised Networks:The Linux Project, in: First Monday, Vol. 6, No. 11, 2001, http://www.first-monday.dk/issues/issue6_11/dafermos/index.html

de Icaza, Miguel (2001): Is Linux ready for the corporate desktop?, in: Computer-world, August 9, 2001, http://www.computerworld.com/storyba/0,4125,NAV47_STO62924,00.html.

Delio, Michelle (2000): It'll Be an Open-Source World, in: wired.com, August 15,2000, http://www.wired.com/news/print/0,1294,38240,00.html.

Dyck, Timothy (2001): MySQL, in: CNET.com, March 5, 2001, http://enterpri-se.cnet.com/enterprise/0-9513-707-6618957.html?tag=st.it.9500-704-0.sr.9513-717-6716336-7053536.

Eduak, Thomas (2001): Nur Fenster nach Windows-Norm? In: ComputerwocheNr. 44/ 2001, p. 38, October 19, 2001.

EITO (2001): European Information Technology Observatory 2001.Evans Data Corporation (2001): Embedded Systems Developer Survey 2001, August

2001, http://www.evansdata.com.Farbey, Barbara; Finkelstein, Anthony (2001): Evaluation in Software Engineering:

ROI, but more than ROI”. Paper submitted for “Third International Work-shop on Economics-Driven Software Engineering Research” (EDSER-32001), http://www.cs.virginia.edu/~sullivan/edser3/finkelstein.pdf.


5.1 References52

Forschungsstelle für Verwaltungsinformatik, Universität Koblenz-Landau (2000):Protokoll der Veranstaltung ‘Open Source Software (OSS) in der Bundesver-waltung’ der Koordinierungs- und Beratungsstelle für Informationstechnik inder Bundesverwaltung, 26.-28. September, http://linux.kbst.bund.de/auftakt/ergebnis/oss-protokoll.pdf.

Freyermuth, Gundolf S. (2001a): Offene Geheimnisse. Aus der Open-Source-Ge-schichte lernen, Teil I, in: c’t 20/2001, p. 176, http://www.heise.de/ct/01/20/176/default.shtml.

Freyermuth, Gundolf S. (2001b): Offene Geheimnisse. Aus der Open-Source-Ge-schichte lernen, Teil II, in: c’t 21/2001, p. 270, http://www.heise.de/ct/01/21/270/default.shtml.

FSF/ GNU: Categories of Free and Non-Free Software, http://www.gnu.org/philo-sophy/categories.html.

Gertz, Winfried (2001): Das Consulting-Geschäft um Open Source, in: Computer-woche, April 27, 2001.

Ghosh, Rishab (1998): Cooking pot markets: an economic model for the trade in freegoods and services on the Internet, in: First Monday, Vol. 3, No. 3, March1998, http://www.firstmonday.org/issues/issue3_3/ghosh/index.html.

Ghosh, Rishab; Prakash, Vipul Ved (2000): The Orbiten Free Software Survey, in:First Monday, Vol. 5, No. 7, July 2000, http://www.firstmonday.dk/issues/issue5_7/ghosh/index.html.

Goder, Andrea (2001a): Entdeckung der Linux-Welt, in: Computerwoche Special,02/2001.

Goder, Andrea (2001b): Pinguine und Pleitegeier: Open-Source-Firmen in der Kon-solidierungsphase. Computerwoche Special, 02/2001.

Hall, Mark (2000): Start-up's Apps Make Desktop Linux Easier, in: Computerworld,September 9, 2000, <http://www.computerworld.com/cwi/story/0%2c1199%2cNAV47_STO63615%2c00.html.

Hecker, Frank (1999): Setting up Shop, http://www.hecker.org/writings/setting-up-shop.html.

Hoch, Detlev J. et al. (1999): Secrets of Software Success: Management Insights from100 Software Firms around the World. Boston, Massachusetts, 1999.

Hoffmann, Naomi (1999): Open Source Software, http://public.kitware.com/VTK/pdf/oss.pdf.

IDC (2000a): Linux: What’s the use? Western Europe 1999-2004.IDC (2000b): IDC IT Forecaster: Computer Operating Environments Coevolve,

August 8th, 2000, http://www.idc.com/itforecaster/itf20000808.stm.Jaeger, Till (2001): Klage wegen GPL-Verletzung, August 13, 2001, http://

www.ifross.de.Jaeger, Till (2000): Copyright oder Copyleft, in: Computerwoche Spezial, 4/2000,

p. 36, , http://www.ifross.de/ifross_html/art6.html.Johnson, Justin Pappas (2000): Some Economics of Open Source Software. Decem-

ber 2000, http://www.idei.asso.fr/English/EPresent/.Kaven, Oliver (2001): Performance Tests: File Server Throughput and Response Ti-

mes, in: PC Magazine, November 13, 2001, http://www.pcmag.com/print_article/0,3048,a%253D16554,00.asp.

KBSt (2000): KBSt-Brief Nr. 2/2000: Open Source Software in der Bundesverwal-tung, http://linux.kbst.bund.de/02-2000/.

Köppen, A.; Nüttgens, M. (2000): Open Source: Strategien für die Beratung. In:Scheer, A.-W.; Köppen, A.: Consulting – Wissen für die Strategie-, Prozess-und IT-Beratung, Saarbrücken, pp. 231-242, http://www.iwi.uni-sb.de/nu-ettgens/Veroef/Artikel/Consulting/OpenSourceStrategien.pdf.


535.1 Literature

Lerner, Josh; Tirole, Jean (2000): The simple Economics of Open Source, December2000, http://www.univ-tlse1.fr/idei/Commun/Articles/Tirole/Simple-Econo-mics.pdf.

Lünendonk (2000): IT-Beratungs- und Systemintegrations-Unternehmen inDeutschland, http://www.luenendonk.de/html/marktanalysen_listen.html.

Miller, Robin (2001): Learning from Mozilla's mistakes. NewsForge, October 23,2001, http://www.newsforge.com/article.pl?sid=01/10/20/1841215&mo-de=thread.

N.N. (2001a): Star Office 6 löst sich vom Microsoft-Vorbild, in: ComputerwocheNr. 42/ 2001, pp. 26-27, October 5, 2001.

N.N. (2001b): Sprung auf den Desktop für Linux zu hoch?, heise online News, May15, 2001, http://www.heise.de/newsticker/result.xhtml?url=/newsticker/data/odi-25.05.01-002/default.shtml&words=Sprung%20Desktop.

Netcraft (2001): Netcraft Web Server Survey, October 2001, http://www.netcraft.com/survey.

Nüttgens, M.; Tesei, E. (2000a): Open Source – Konzept, Communities und Insti-tutionen, in: Scheer, A.-W. (Editor): Veröffentlichungen des Instituts fürWirtschaftsinformatik, Heft 156, Saarbrücken, http://www.iwi.uni-sb.de/nu-ettgens/Veroef/Artikel/heft156/heft156.pdf.

Nüttgens, M.; Tesei, E. (2000b): Open Source – Marktmodelle und Netzwerke, in:Scheer, A.-W. (Editor): Veröffentlichungen des Instituts für Wirtschaftsinfor-matik, Heft 158, Saarbrücken, http://www.iwi.uni-sb.de/nuettgens/Veroef/Artikel/heft158/heft158.pdf.

Nüttgens, M.; Tesei, E. (2000c): Open Source – Produktion, Organisation und Li-zenzen, in: Scheer, A.-W. (Editor): Veröffentlichungen des Instituts für Wirt-schaftsinformatik, Heft 157, Saarbrücken, http://www.iwi.uni-sb.de/nuettgens/Veroef/Artikel/heft157/heft157.pdf.

O‘Reilly (1999): Open Source; kurz & gut, Köln.Open Source Initiative: The Halloween Documents, http://www.opensource.org/

halloween/.Osterberg, Jürgen (2001): Flotter Auftritt mit Zope, in: Computerwoche Nr. 44/

2001, p.26, October 19.Rasch, Chris (2001): The Wall Street Performer Protocol: Using Software Comple-

tion Bonds to Fund Open Source Software Development, in: First Monday,Vol. 6, No. 6, June, http://www.firstmonday.dk/issues/issue6_6/rasch/.

Raymond, Eric (1997): The Cathedral and the Bazaar, http://www.tuxedo.org/~esr/writings/cathedral-bazaar/.

Raymond, Eric (1998): Homesteading the Noosphere, http://www.tuxedo.org/~esr/writings/cathedral-bazaar/.

Raymond, Eric (1999): The Magic Cauldron, http://www.tuxedo.org/~esr/writings/cathedral-bazaar/.

Rosenberg, Donald K. (2000): Open Source: The Unauthorized White Papers, Fos-ter City.

Sandred, Jan (2001): Managing Open Source Projects: A Wiley Tech Brief.Schenk, Thomas: Linux: Its history and current distributions, http://www.develo-

per.ibm.com/library/articles/schenk1.html.Schmitz, Ludger (2001): The use of open source software in the public sector, http:/

/europa.eu.int/ISPO/ida/?http&&&ag.idaprog.org/Indis35prod/doc/333.Security Space (2001): Security Space Web Server Survey, October 2001, http://

www.securityspace.com/s_survey/data/200110/index.html.Succi, Giancarlo; Paulson, James; Eberlein, Armin (2001): Preliminary Results from

an Empirical Study on the Growth of Open Source and Commercial Software


5.2 References54

Products. Paper submitted for “Third International Workshop on Economics-Driven Software Engineering Research” (EDSER-3 2001), http://www.cs.vir-ginia.edu/~sullivan/edser3/paulson.pdf.

Symons, Charles (2001): Controlling software contracts, http://www.gifpa.co.uk/li-brary/Papers/Symons/esepg97.html.

TechConsult (2001): Fast jedes 5. Unternehmen setzt 2003 auf Linux. Press Release,http://www.techconsult.de/de/Services/studiendetail.cfm?id=82.

Tuomi, Ilkka (2001): Internet, Innovation and Open Source: Actors in the Network,in: First Monday, Vol. 6, No. 1, January 2001, http://www.firstmonday.dk/issues/issue6_1/tuomi/index.html.

Venture Development Corporation (2001): Linux's Future in the Embedded Sys-tems Market (Summary for the Embedded Linux Community), June, http://www.linuxdevices.com/articles.

Wayner, Peter (2001): Open source databases bloom, in: Computerworld, Septem-ber 9, 2001, http://www.computerworld.com/storyba/0,4125,NAV47_STO63629,00.html.

Weiss, George (2001): The Future of Linux and Open Source. Gartner, Strategy &Tactics/Trends & Direction, Note Number AV-13-9850, June 20, 2001.

Wheeler, David (2001): More Than a Gigabuck: Estimating GNU/Linux's Size, Ver-sion 1.06, 30.6.2001, updated 8.11.2001, http://www.dwheeler.com/sloc/redhat71-v1/redhat71sloc.html.

Wheeler, David (2001): Why Open Source Software / Free Software (OSS/FS)? Lookat the Numbers!, http://www.dwheeler.com/oss_fs_why.html.

Whitlock, Nathalie Walker (2001): The security implications of OSS, March 2001,http://www-106.ibm.com/developerworks/linux/library/l-oss.html?open&l=252,t=gr,p=SecImpOS.

Wiehr, Hartmut; Wild, Martin (2001): Linux-Die Alternative, in: Informationweek,July 7, 2001, http://www.informationweek.de.

Wysong, Thom (2000): Introduction to Open Source and Free Software, http://www.technodemocracy.org/papers/ossfs.html.

Yager, Tom (2000): BSD's strength lies in devilish details. InfoWorld, November 2,2001, http://www.infoworld.com/articles/tc/xml/01/11/05/011105tcbsd.xml

5.2 Web resources

5.2.1 Publications on OSS and Linux

Daemonnews.org(monthly, BSD) http://www.daemonnews.org/Enterprise Linux today http://eltoday.com/Linux Community.de http://www.linux-community.de/Linux Weekly News http://lwn.netLinux World http://www.linuxworld.com/Linux.com http://www.linux.com/LinuxBusiness.com http://www.linuxbusiness.com/LinuxInsider.com http://www.LinuxInsider.com/Linuxjournal http://www2.linuxjournal.com/index.htmlLinux-Magazin http://www.linux-magazin.de/LinuxNews.com http://www.linuxnews.com/Linuxtoday.com http://linuxtoday.com/


555.2 Web resources

LinuxUser http://www.linux-user.de/Newsforge http://www.newsforge.com/?578Pro-Linux http://www.pro-linux.de/Slashdot http://slashdot.org/Computerwoche http://www.computerwoche.dect, iX http://www.heise.de

5.2.2 Publications on IT in general

ZDNet http://www.zdnet.comInfoworld http://www.infoworld.comCIO.COM http://www.cio.comDarwin http://www.darwinmag.com/Computerworld http://www.computerworld.comIDG.net http://idg.net/InformationWeek http://www.informationweek.comIT world http://www.itworld.com/WSTA Cyberlibrary http://cyberlibrary.wsta.orgOS opinion http://www.osopinion.com/NetworkWorld http://www.networkworld.comIT papers http://www.itpapers.comBitpipe http://www.bitpipe.com/

5.2.3 Software aggregation

Freshmeat http://freshmeat.net/Linux Center http://www.portalux.com/Linux.org http://www.linux.orgLinuxApps.com http://www.linuxapps.comLinuxISO http://www.linuxiso.org/Tuxfinder http://tuxfinder.com/

5.2.4 Organisations/OSS community

BerliOS http://www.berlios.deFree Software Foundation http://www.fsf.orgGNU.org/ FSF http://www.gnu.orgLinux International http://www.li.org/Linux Professional Institute http://www.lpi.orgLinux Verband e.V. http://www.linux-verband.de/Linux.org http://www.linux.orgOpen Source Development Lab http://www.osdlab.org/Open Source Initiative (OSI) http://www.opensource.orgOSDN.org http://www.osdn.com/Silicon Valley Linux User Group http://www.svlug.org/Sourceforge http://sf.net


5.2 References56

5.2.5 Commercial OSS companies

ActiveState http://www.activestate.comLinux Global Partners http://www.linuxglobalpartners.com/Linux Information Systems AG http://www.linux-ag.deLinuxCare http://www.linuxcare.comLinuxMall.com http://www.linuxmall.com/LinuxTshirts.com http://www.linuxtshirts.com/MandrakeSoft http://www.linux-mandrake.comMySQL AB http://www.mysql.comNuSphere http://www.nusphere.com/O´Reilly http://www.oreillynet.com/Penguin Computing http://www.penguincomputing.comPostgreSQL, Inc. http://www.pgsql.com/QliTech http://www.qlilinux.comRed Hat http://www.redhat.comSendmail, Inc. http://www.sendmail.comSuSE http://www.suse.deTrolltech http://www.trolltech.comVA Linux http://www.valinux.comZope Corporation(formerly Digital Creations) http://www.zope.com

5.2.6 OSS projects and products

AbiWord http://www.abiword.orgApache http://www.apache.orgBind http://www.isc.org/products/BIND/CVS http://www.cvshome.orgEnhydra http://www.enhydra.orgFreeBSD http://www.freebsd.orgGimp http://www.gimp.orgGNOME http://www.gnome.orgGNU Project http://www.gnu.org/software/software.htmlGnu.org http://www.gnu.orgGPG http://www.gnupg.orgInterbase http://www.borland.com/interbaseKDE http://www.kde.orgKoffice http://www.koffice.orgLinux http://www.linuxhq.orgMailman http://www.list.orgMesa http://www.mesa3d.orgMozilla http://www.mozilla.orgMySQL http://www.mysql.comOpenOffice http://www.openoffice.org,Perl http://www.perl.comPHP http://www.php.netPostgreSQL http://www.postgresql.orgPython http://www.python.orgQt http://www.trolltech.comSamba http://www.samba.org


575.2 Web resources

Sendmail http://www.sendmail.orgStaroffice http://www.sun.com/starofficeTcl http://www.scriptics.com/Wine http://www.winehq.orgZope http://www.zope.org


5.2 References58
