Paper Q

COMPARATIVE CLASSIFICATION

DEPTH CLASSIFICATION 35

Gives an account of comparative studies in thefield of documentation. Describes the recentcontributions on general theory of classifica-tion by various schools. Indicates the workdone on coding for machine search and bringsout its limitations. Mentions the comparativestudy now being conducted at the College ofAeronautics at Cranfield and outlines thefuture lines of development.

1 COMPARATIVE STUDIES

When it becorn es possible to speak of usingthe comparative method in the study of a sub-ject, it is a sign that the subject has attained acertain maturity _ one might even say respecta-bility. For the essence of the method is that itstudies side-by-side the ways in which the sub-ject has been approached and developed indifferent sets of circumstances; and the wish tocompare implies two assumptions; that thedifferent ways have SOIne comrnon features, andthat each way has sorneth irig of value to contri-bute to the general study of the subject. Insofaras the comparisons are also usually based on thefoundations laid in different countries, it maytruly be said that the use of the method is an indication of the growing desire among scholars tobreak down unnecessary national barriers and todraw strength from each other's work.

V 6 N 4 Dee 7959

..

D J FOSKETTI nst I t ut. of Educot Ion. Unl ver sl t y of London

11 Studies in Other Disciplines

Several disciplines have already benefited inthis way: religion and education are w el l s knownexamples, and they demonstrate the range ofpossible comparisons. In the first, it has beenthe philosophy, the theory so to speak, that hasprovided most of the inspiration. There have,it is true, been comparative studies of religiousc er ernoni ee and rituals, but the greater interesthas always attached to similarities and differ-ences in the mystique, the sacred books andtheir int erp r et at Ion. In education, on the otherhand, greater attention is given to the practicalmanifestations. The doctrines of many of th eeducational philosophers have transcended theboundaries of their country or culture and havewon international acceptance, but differenceshave arisen in the lines along which the p r ac ti ,cal applications of their principles have de-veloped.

12 Studies in the Field ofDocumentation

The field of documentation already providesu~ with SOIne examples of both types of app-roach. Perhaps the outstanding example is thatof Dr S R Rarig anat han on cataloguing, published

105

012 FOSKETT

in :1955 with the title Heading and canons: com-parative study of five cataloguing codes. Inclas sification, the International Federation forDocumentation some years ago set up its FID/CACommittee, which has produced several reports,some of them comparative, but it would be toomuch to say that comparative classification hasyet won an established place in the study of thescience. An examination of the work that hasbeen done shows that, although much of the pre-liminary surveying is complete, little in theway of a systematic foundation has been laid,and we do not have to search very far to dis-cover the extent to which classification hassuffered from this lack. We in the West, forexample, know next to nothing of the new sy s ,tem that has been made in the USSR on the basisof the philosophy of dialectrical materialism.

:13 Descriptive Surveys

What we can say, however, is that the des-criptive surveys have been made with very greatthoroughness. Without going too far back intohistory, we can point to the books of Flint:Philosophy as scientia sc ient.ar urn, and a historyof classifications of the sciences :1904; of Bliss:The organisation of knowledge and the system ofthe sciences, 1929; and of Sayers: A manual ofclas sification for librarians and bibliographers,third edition 1955. All of these adm irable andencyclopedic works contain descriptions ofmany schemes, but 1 believe that their authorswould agree with me when 1 say that they weremore concerned with recording the theory andhistory of the science than with developing thecomparative study. If we are able to see far-ther today, We can with gratitude, like Newton,admit that we stand on the shoulders of giants.

14 Systematic Study

Some steps forward have been taken. Inclas sification, as in cataloguing, a sys tematiccomparative study has been made by Dr Ranga-nathan in the new edition of his Prolegomena tolibrary classification, 1957; this is a goodexample of genuine comparative classification,in that it analyses in detail the methods adoptedby several well-known schemes of classificationin solving certain problems, and uses the re-sults of the study to make improvements to the

106

Colon Classification itself. Ra nganath an isoften accused in the West(usually by criticsignorant of his writings) of being over-theo-retical, and of allowing a mystical p r e c occup a ;tion with eternal truths to blind him to thepractical problems of classifying documents.But like Plato, who also exhibits a certain pre-occupation with eternal truths, Rarig an ath ant smain characteristic is a bus ines s -like deter-mination not only to come to grips with diffi-culties, but also to propose feasible solutions.Nowhere is this demonstrated more clearlythan in Part 7 of Prolegomena.

2 GENERAL THEORY

21 Ra ngana thant s Contribution

As is well known, Ra nganath an aims alwaysat a general theory of clas sification, and hiscomparative studies have contributed to thisend, but his hypotheses are always subjectedto the test of incorporation into the ColonClas sification and application to the arrange-ment of documents. While admitting withoutreserve the correctness of this method, wemust nevertheless recognise that it has herethe limitation that Colon itself is so incompletein many important details. T H Huxley onc esaid that Herbert Spencer's idea of tragedy w a s"a beautiful hypothesis slain by an ugly fact",and 1 occasionally wonder whether Ran gariarhanprefers to avoid such tragedies by using theflexibility of Colon to gloss over incompatibi-lities.

22 British Classification ResearchGroup

Two other writers who have avoided such alimitation because they have not been concernedto improve a pre-existing system are B C Vic-kery and Eric de Grolier. Vickery's book,Classification and indexing in science, appearedonly las t year, but the second edition is alreadyin the pres s and is expected shortly. It re-presents the results of several years of inten-sive study and discussion of basic principles andtheir applications made by the British Classifi-cation Research Group, of which Vickery hasheen prime mover and secretary since itsformation. This group began its work as a

An lib sc

COMPARATIVE CLASSIFICATION

consequence of the Royal Society ScientificInformation Conference of 1948, which re-commended, largely owing to the initiative ofProfessor J D Bernal, that the study of cl a s s i ;fication should be und er tak en by interestedpeople. It has no official ex i st en c e, but itsmembers have met regularly, and have bo m ,barded each other with reports and papers,most of them unpublished. Reports of itsprogress appear from time to time in As l ib+sJournal of documentation. Its members, bet-ween them, have during these years compiledsome twenty spec ial i s t classification systems,several of them quoted in Vickery's book.

23 Vickery's Contribution

The chief value of the book (apart from. itsexcellent presentation of the English work) isthat Vickery set out to make a comparativestudy of principles - whether these principleswere explicitly stated by their users or not -and extended the range of his survey to in-cIu de the new g'roup s working on machineliterature searching and machine translating.He was able to show' how the various schoolsof thought were tending tv arrive at very simi-lar results, even though they might havestarted from Widely separated bases. It is infact a feature of Vickery's work that he con-stantly seeks to underline what is common.rather than what is distinct. in the differenttheories, and he is probably doing more thananyone else to engineer fruitful contacts bet-ween workers in different count r ie s .

24 De Gz-ol ier !s Contribution

I hope he will agree with me when I saythat this is also the aim of Dr Eric de Gr ol ie r ,whose encyclopedic work. Theorie et pratiquedes classifications docu rnent air e s , appearedin 1956. the result. he tells us, of more thantwenty years work. De Grolier has been asso-ciated with Rang anathan for several years asco c r app or t eu i- to the FID committee on theGeneal theory of c Ia s s if ic t ion , and we havebeen glad (and even at times amused) to seehow these two original minds have strucksparks from each other. even if it is some-times in self-defence!

V 6 N 4 Dee 1959

Q3

In this. and in a report written for Unesco,the range of de Gr ol ier ts work is formidable.He is completely familiar with recent thought.not only in Western Europe. Great Britain andthe USA. but also in Eastern Europe, theUSSR. and China. So far as I know. the com-bination of this wide range with a penetratinganalysis oi principles is unique in classifica-tion literature. But while his main aim in thebook is analytical and descriptive, he doesgive an outline of his own Ralpha-synthetique"system, which he abbreviates to the conve-nient name of ALSYN. I must admit that Ihave found it a little hard to apply thisscheme, but this is probably because I haveseen only the outline, which to me is charac-terised by what the English think to be typicalof the genius of the French people, a reductionto a closely-woven set of logical abstractions.

3 CODING FOR MACHINE SEARCH

All these three writers have advanced beyondtheir predecessors in one very important parti-cular: they take into account the vast amountof work now being carried out on the principlesof coding for searching the literature by ma-chines. In his paper for Unesco in 1950.Classification, coding and machinery forsearch, Ranganathan took up the position whichhe still maintains, that the much-praised effi-ciency of machine searching has not yet beenproved. and that some kind of synthetic classi-fication will be in any case forced on themakers of machine codes if they are to avoidthe immense searching labours that are madeinevitable by random codes. The same view,generally speaking, has been adopted byVickery. though he differs on some points. Heand de GroHer have discussed in detail thework of teams dealing w itt. machine coding forinformation retrieval and for mechanical trans-1ation. It has Been suggested that the rele-vance of the latter work makes it useful toadopt the term. "codification" instead of"classification". which is held to be narrowerin meaning. Without entering into <1 contro-versy over terms. I shall continue here touse the term preferred in England. "classi-ficatlOn". while readily acknowledging thenecessity for widening its meaning beyondthe mere arrangement of books on libraryshelves.

107

Q3l

31 Limitations of MechanicalRetrieval Systems

Vickery's latest papers, I am glad to say,contain sorn e sharp criticisms of the machinetheorists. They were presented to the recentconferences in Paris and Cleveland and arenot yet published, but they incline to the view,with which I am in full sympathy, that wheremachines are wanted is in putting informationinto retrieval systems, and not in getting itout of them, as has always been taken forgranted. Once a document or piece of infor-mation has been coded, its filing into a sv s ;tematically arranged collection of otherdocuments is a purely clerical operation;whereas the essence of getting informationout of the system is the establishment ofrelevance _ the relevance of the informationscanned to the information sought. I do notsee how, for this operation, machines canever be expected to replace the skill andknowledge of an experienced librarian ordo c u rne nt a l is t ,

Another criticism with reference to thepresent situation only, appears in the papersubmitted to the Cleveland conference byDr de Grolier. This paper examines brieflythe several contemporary schools of thoughtengaged on designing coding systems, andmakes a welcome plea for greater coIl abo r a ;tion and the breaking down of prejudices. Iwas disappointed only in that it did not gofurther and give more precise details of whatlines de Grolier himself thinks should be ex-plored. Let us hope that the Cleveland con-ference will be somewhat more productive ofuseful proposals than was the Washington con-ference.

4 CRANFIELD EXPERIMENT

So much for a necessarily brief glance atrecent comparative studies of principles andtheories. An important piece of practicalresearch is now being conducted in Englandat the College of Aeronautics at Cranfield.Asl ib has organised a test which will consistof indexing by four different indexing systems,some 20, 000 documents dealing with aero-nautical subjects, and then testing the effi-ciency of the systems in answering actual en-

108

FOSKETT

qu i r i es , One of the systems is a facetedclassification produced in a matter of only afew months by Vickery and J E Farradane,and the March, 1959, issue of th e Journal ofdocumentation contains Bulletin 5 of theClassification Research Group, a discussionof some practical problems arising out of theuse of this system. Its value is proved to meby the direct relevance of the discussion tosome work of nly own on a system for docu ;mentation in the field of Occupational Safetyand Health.

We can see, then, that although comp a r a-,tive classification n12Y be in its infency as arecognised branch of classificatory science,nevertheless sufficient work has been done onboth theoretical and practical sides to allowus to see the value, indeed the necessity, ofusing this method of study. I should now liketo put forward some suggestions for futurelines of development which I think mightprofitably be pursued.

5 FUTURE LINES OF DEVELOPMENT

Perhaps the most obvious of these appearsto be a comparative study of the arrangementby existing schemes of a selected number ofdocurnent s , in the manner of Ra.ngan a than t sHeading and canons for cataloguing codes.This is certainly what is being done atCranfield, but there a single subject field waschos en and each of the systems had to bedeveloped more or less for the purpose of thisparticular project. What I have in mind nowis the cornpar i son, over a much wider field,of what system-makers have already donethemselves to cope with everyday libraryproblems.

51 Wider Coverage

This would be a rather elementary exercise,however, and I have some doubts as to whetherit would in fact be worth doing. In my ex-perience, the feeling of dissatisfaction withexisting schemes is so widespread that wemight be justified in assuming that such asurvey would tell us very little more than weknow already. On the other hand, theseschemes have been built up on a bssis of pr a ct i,

An lib sc

COMPARATIVE CLASSIFICATION

cal exp erienc e, ana some at leas t of theirmethods deserve consideration. Vickery, forexample, in Appendix C of his book, comparesthe categories actually used in the natural sci-ences in Colon and UDC; and in my chapter onClassification in the Aslib Handbook of speciallibrarianship 1 gave brief treatment to thevarious methods of revision. It might thereforebe useful to look at this kind of specific feature,as they seem likely to have to be incorporatedinto any future scheme. More, 1 feel, mightbe made in this way of the experience of In sd oc ,where they have been cla s s ify ing the samearticles by Colon and UDC for sometime. Twopoints that strike one immediately in the Insdoclist are the greater simplicity of Colon's nota-tion, and the lack of detail in its schedules,which is due. 1 am told, to their classifyingwith the title only before the text is released.

52 Basic Classes

One of the major problems to be faced byany system-maker is the recognition of whatare traditionally called "basic classes", andthe influenc e of the completely synthetic app-roach exemplified in some machine systems,especially those that rely on purely randomcoding, has tended to conc eal the importanceof this. Whatever we call them, "subjects","fields of study", "basic c las s e s ", these par-tially-comprehensive groups are absolutelynecessary to any general system of classifi-cation. Indeed, even in making a specialistsystem, where one begins by naming the sub ,ject one is going to classify, one cannot escapethe problem, because writers will producedocuments that cover several of what appearin the schedules as separate c12s s es. Togive one example only, in Occupational Safetywe have the separate Fa.za r ds "Fires", "Ex_plosions" and others; but we must also havea partially-comprehensive heading "Fires andexplosions" because they are often treatedtogether in the same document.

53 Basis of Enumeration

So far, the naming of basic classes hasbeen a more or less subjective process; Blisscertainly referred to the educational and sci-entific consensus, but it is well known that

V 6 N 4 0 Dee 1959

Q54

there is no unanimity whatsoever about thisconsensus. The Classification Research Grouphas recently begun to give some attention tothe problem, and is investigating the conceptof "levels of integration" to see whether itwould provide an acceptable scienti.fic basisfor deciding on the choice of main classes.Professor B M Kedrov of the Moscow Academyof Social Sciences, in a paper presented to theSecond International Conference on the Philoso-phy of Science in Zurich in 1954, and elaboratedin Voprosy pilosofii in 1955, also studied theproblem of reducing a ramified classificationof the natural and social sciences to the lineartype required for arranging documents. Hisclas sification was founded on dialectical mate-rialism, and contained several features thatclos ely re sembled the outline sys tem resultingfrom the use of the integrative levels. Col.Ia ,boration on the selection of basic classes whichmust probably be postulates afresh for eachepoch as a preliminary step to the constructionof a new general scheme of classification, there-fore, shows every sign of being desirable anduseful.

54 Special Classifications

With only a few exceptions, the specialistsystems that have recently appeared in suchnumbers have been worked out more or less inisolation. It is true that many of them haveused the techniques of general schemes, butthey have not been compiled with a view to in-corporation in the general schemes. All theschemes made by members of the ClassificationResearch Group, for example, have usedRanganathan's technique of facet analysis, butonly two of them, for Management and for FoodTechnology, were intended to be expansions of theColon scheme. An immensely detailed classi-fication for Agriculture has also been producedas an elaboration of Clas s J in Colon, by Dr D BKrishna Rao in India. This is probably the bestexample to date of a faceted scheme, of theoriginal Colon type, made for a special subject.Lately, another exercise in integration hastaken place: the system made for Management,and another being made for Community Develop-ment, have been examined as possible compo-nents of the Social Science Classification madeby Miss, Barbara Kyle for use in the UnescoSocial Sc ienc e bibliographies.

109

Q55 FOSKETT

55 Experience of UDC

By far the greatest amount of experience inexpanding special subjects within a generalframework has of course been gain.,d in thedevelopment of UDC, and the influence ofmodern theories, in particular that of facetanalysis, is obvious in some of the latest pub-lications - in 669 Metallurgy, and in 621. 039Nuclear Technology, for example. Dr DonkerDuyvis, and Mr G Lloyd of the British Stan-dards Institution, have both expressed the viewthat UDC ought to take new thought into account,and that its structure is sufficiently flexible toallow the incorporation of new techniques ofconstruction.

6 COMPARATIVE CLASSIFICATION

What particularly interests us here, how-ever, is not so much the process of integration,important as that is, as the comparative studyof the techniques of making general and specialsystems. It may well be tnat t ht s should formthe core of comparative classification, becausethe specialist systems are likely to provide thetesting grounds for new theories. Less labouris involved, while there is a t the same timegreater pressure on us to produce systems tomeet specific needs; all the CRG systems,for example, have been made to order, inresponse to the wish of subject specialists.We have not yet answered the question as towhether the techniques that prove satisfactoryfor specialist systems will be useful in makinga general system. Can we fo r rrrul at e generalprinciples for choice of facet sequence? If theanswer is yes, then I can see no reason formaintaining the view that the requirements ofthe two types of system are incompatible. Thisview is held, I should say, by the majority ofspecialist workers today.

61 Notation

When we turn to no ta tion, we find an almo s tcomplete lack of ev.idence that would enable usto choose a satisfactory form of code. Manypronouncements have been made, no doubt, anda great deal of work published. But the prob-lem that is unresolved in my mind, and I sus-pect in the minds of many others, involves the

110

psyc nokog ic al acceptability of different kinds ofnotation. It is obvious, for example, that themore species of symbols one is prepared touse, the shorter will be the length of t he nota-tion of complex subjects; and we all know thatone of the strongest and most persistent criti-cisms of UDC symbols is their extreme length.Ranganathan has demonstrated many times thegreat superiority of Colon in this respect. Yetmy own experience leads me to think that themost important single features of notation in aclassification system although not for machinecoding is that it should convey clearly a recog-nisable and familiar sequence, and that usersare content to accept longer symbols for theirsubjects because they are not repelled by theirappearance. The notation, in fact, should notinterpose an unnecessary degree of internalcomplexity. But it may be that this considera-tion for the aesthetic s of notation must besacrificed in this technological age, like somany other aesthetic considerations.

7 NEED FOR HUMANINTERPRET ATlON

It will have become obvious by these re-marks on notation, if it. was not already obviousbefore, that I am implicitly taking for grantedthe presence in the information retrieval sys-tern of a human operator - manifestly a largeassumption today. I am quite happy to becalled medieval, scholastic, reactionary, orany other adjective suitable to apply to thosewho wish to hold up the march of mechanisation.I am unrepentant, and I stand by the words ofRanganathan some years ago: "At the finalstage, documentation service to living menshould be done by living men". But a great dealof our comparative studies will be concernedwith machine techniques, for it is imperativethat we should discover the power and the limi-tations of machines. The matter of notationpresents an interesting example of completecontrast. All kinds of details have to be takeninto account when planning for a human opera-tor, and I have just said that w; know very littleabout the factor of acceptability; the machine,on the other hand, has little aesthetic sensiti-vity, and is happiest with a binary notationwhich would be almost incomprehensible to ahuman.

An lib sc

COMPARATIVE CLASSIFICATION

72 Machine Performance

It would be quite superfluous to embarkhere upon a description of any of the multitudeof machines whose proud owners already fillfat volumes with their praises. 1 do not conc ea lmy astonishment and ad rnir at ion at their perfor-mances. 1 can find only one fault, common toall, or nearly all: that they were not designedfor information retrieval, and consequently theretrieval system has to be suited to the machineand not vice versa. 1 do not doubt for onemoment that a jet engine would be capable ofdriving my lawn mower, but it is scarcely anappropriate match on either side. If it is true,as 1 have heard, that our documentation col-leagues in the USA are beginning to be put underpressure from the management to adapt theirsystems to suit the firm's computer, this is amost depressing situation which is bound to havea bad effect on their theoretical and practicalwork.

73 Mechanical Translations

The latest analogy to be drawn in the sphereof machine theory is with mechanical translation.It is clear that a classification scheme is a kindof artifical language. S C Bradford, of theLondon Science Museum Library, who wasperhaps most responsible for the success ofUDC in England, used to believe that UDCcould act as an international dictionary; andcertainly the recent trilingual edition servesthis purpose to a quite considerable extent.Dr. Ranganathan has long used the phrase,"an artificial language of ordinal members" todescribe a classification scheme.

74 Limitations

It is this specific relationship that has beenthe subject of the conference just concluded atWestern Reserve University in Cleveland:"Standards on a common language for machinesearching and translation ". It is hard for meto be very optimistic about the results of sucha discussion, because 1 have so many doubtsabout the appropriateness of machines for lit-erature searching, but 1 certainly share thehope expressed by Dr de Grolier in the paperhe submitted - that at least the conference

V 6 N 4 Dee 1959

Q75

would succeed in drawing more closely togetherthe efforts of the various people who are work-ing on problems of classification. First re-ports seem to show that the conference was infact a greater success than 1 had thoughtpossible.

75 Need for Objectives

But it is necessary for us to be clear aboutour objectives before becoming enthusiasticabout new techniques, and there are some verypertinent differences between translation andinformation retrieval. In translation, one isdealing with a text of specific known content.In the proces s of matching one language withanother, the limits of expression and info r m a,tion are quite clearly defined, and leaving asidenuances of style, one can say that the productionof a satisfactory "approximate" translation canbe reduced to a mechanical operation becausethe limits are definite and known. This is notthe case in information retrieval. Very often,the expression of the question is vague andimprecise, but even if it is not, the chancesare that the limits of the answer cannot begiven with precision. It is obviously hard foran enquirer to be quite definite about what hedoes not know. As is often said, then, thecentral problem for information retrieval isthe detection of relevance - even, in a well-documented subject, of minute degrees of rele-vance. The greater the skill of the documenta-list, the less will be the mass of literaturethat has to be sorted by the enquirer himself.No doubt a mechanical search can 'producebibliographies and abstracts; but all myown experience tells me that it is the skilland flair of the documentalist that really savethe time of the research worker. 1 know thatProfessor Bernal, for example, reported atWashington that the output of scientific litera-ture has reached such proportions that manyscientists now prefer to return to a seven-teenth centruy technique and get their infor-mation by direct contact. 1 know, too, thatthis is a very obvious and efficient technique;1 often use it myself. But 1 believe that toadopt this technique as a regular system wouldvery soon impose an intolerable burden, onethat a scientist accustomed to a good informa-tion service from a specialist library wouldflatly refuse to accept.

111

Q76 FOSKETT

76 Lack of Proper Comparison

One of the most important tasks facing us,therefore, is a detailed s ta t ern en t of the needs.and processes that arise in an informati.on re-trieval system and a mechanical t r ansl at rngcentre. Some years. in an article in Americandocumentation, Mr Vickery showed quite clearlythat most of the claims made for the superiorityof machine literature searching over conven-tional manual methods could be dismissed be-cause important parts of the retrieval processwere being left out of the comparison. So faras 1 know, this ar gurne nt has never been refuted.

77 Need for F'Inanc ral Support

It has, however, beer, completely ignored,and one of the most disheartening features ofthe present situation is that nearly all existingcomparative studies have been made withoutany unancial backing. On the other hand, moneyhas been poured out with the utmost prodigalityfor work with machines _ and most of the wor k ,er s on th ern go on ignoring the immense pro-gress made over the last few years In both thetheory of c La s sif ica t i.on and also in its applica ..tion in working int~rmation sys tern s. Thiswas well illustrated in the Washington Confer-ence of 1958: an immense amount of time,labour and money went into this prodigiousexercise - aT''' to what end? "Parturientmontes, nascetur ridiculus rrius t",

78 Role of FID

Much of the impetus for these gre<l.t advan-ces in classification has C"TY\ef r orn the FID.If only a fraction of one per cent of the re-sources squandered on machinery had beenavailable for this work, I have little doubtthat we could have produced something muchmore spectacular. It may be one of the mostdifficult things in the world to prove that "-straight line is usually the shortest distancebetween two points; but when we see our

American friends spending several years onproving mathematically that a hierarchicalstructure is sometimes useful in coding be-cause class headings include their sub-divisions, we begin to wonder whether suchpursuit of ultimate truth i« really the bestcourse of action just at this time. Would itnot be better simply tc ca II it a postulate?

8 CONCLUSION

We can all appreciate the value of fastmachine searching in a large centre where theobjectives of the search can be limited andmore or less exactly specified as, for example,in seeking particular data, properties. and soon. But in my view the future of classificationand coding study must be related to actualworking conditions, and it would be great foolish-ness to Ignore the unique contribution made bythe human operator in the type of retrieval sys-tem that is most commonly found in whichmformation, not data, is the objective. Al-though the Washington and Paris conferencesoffer few encouraging signs of an appreciationof this among machine theorists, neverthelesssome of the personal cont ar t s made there, atthe recent Cleveland conference, and aboveall at the Dorking conference in 1957 (a reallyproductive conference): do indicate that morehelpful collaboration is possible. The syste-matic study of comparative classification andcodification is the only sure way of bringingout the advantages and disadvantages of all t hemany systems now being developed. Withoutnecessarily committing ourselves to a searchfor the ideal, universal system, we shouldsurely be able to formulate some postulatesbasic principles and having a wide validity,and practical means of putting them intooperation in different circumstances. Hereis work that is most appropriate for an inter-national body. I am confident that Une sc ocould be persuaded to take an interest, and itwould be most fitting and profitable for FID tocontinue its work on classification along theselines.

112 An lib sc