the brown internet computing lab
TRANSCRIPT
conduit!conduitFall, 2003Volume 12, Number 2 Department of Computer Science
Brown University
25th ANNIVERSARY 79-0
4
Bro
wn
University Computer Scien
ce
THE BROWN INTERNETCOMPUTING LAB
Steve Reiss
ern research must be driven d shown to work in the real
ld on tomorrow’s problems
OverviewWith the support of anew $640,000 two-year grant from theNational ScienceFoundation, the Com-puter Science Depart-ment is developing aninnovative laboratoryfor undergraduate andgraduate educationand research intoInternet-related prob-
lems. The lab will ensure that our research andour students address issues of both immediateand long-term relevance as we move to an Inter-net-based world.
The Proposed LaboratoryThe Internet has changed the face of computerscience. Concepts such as e-commerce, grid com-puting, peer-to-peer programming, andfile sharing provide new directions. Oldproblems, such as security and authentica-tion, have become more complex and atthe same time more important. Otherproblems, such as cooperative work, data-bases, and distributed systems, havechanged scale so dramatically that onenow must talk, say, of millions of concurrentusers. It is difficult to study or teach this new face ofcomputer science using old tools and environ-ments. Many of the complexities and problemsoccur only when tackled at “Internet scale”, thatis, when there are thousands of simultaneoususers, millions of potential users, widely distrib-
Modby anwor
Brown University, Box 1910,
uted systems, multiple servers, unreliable net-works, and strong security and privacy concerns,hackers, and the like. Modern research must bedriven by and shown to work in the real world ontomorrow’s problems. Moreover, students canencounter and appreciate the difficulties inherentin modern Internet-based systems only byworking with real systems.Our new laboratory for experimenting with Inter-net-scale applications will be suitable for bothstate-of-the-art research and instruction. For re-search, it will provide a framework in which ideascan be tested at Internet scale. Such a laboratorywill give researchers insights into what the realproblems are and how they might be addressed,and will also allow them to validate new ap-proaches. For teaching, the laboratory will showstudents the problems that arise and techniquesused when building applications at Internet scalesand will let us offer courses that better preparestudents for real-world programming andresearch.This laboratory will use a flexible combination ofhigh-end servers, large data storage, clustermachines, workstations, network infrastructure,and appropriate software. The center of the con-figuration will be servers capable of runningInternet-scale applications that will run commer-
cial web software, a commercial-scale databasesystem, and load generation and analysis soft-ware. The data storage will support this applica-tion and provide a large web snapshot forteaching and research involving accessing andunderstanding the structure of the web. A labora-tory-based front end of 20 workstations willprovide class and researcher access.
Providence, RI 02912, USA
LoadGeneration
ClusterServer
Front EndLaboratory
Database
In
The lab can be u he clusters to generate load servers to
Filer
The laboratory will be used in a variety ofresearch projects. Pervasive programminginvolves building a common programmingframework for dealing with web services, peer-to-peer computing, and grid computing in anopen-source environment. Aurora*1 tackles theproblems of large-scale network and stream-based data management. Language researchaddresses the semantics of web applications.Covalent objects provide a means for objectsharing at Internet scales. Prooflets are a simpleand efficient way to authenticate large numbersof small data items across the web. Web model-ing will look at statistical models of the web and
their application to search, data mining, annota-tion, and crawling. In addition, we will use thefacility as a laboratory in both undergraduate andgraduate courses.
ConfigurationThe laboratory will have a flexible configurationsimilar to that shown above. It will need: • An SMP Server. This should be a 64-bit, mul-
tiprocessor machine with significant memory.We are interested in having at least eight pro-cessors and 16Gb of memory.
• Cluster Server. This should be a rack-mounted cluster of machines. Our initialmodel of such a cluster would consist of 40dual-processor machines, each with at least2G of memory.
• Load Generator. This should also be a rack-mounted cluster of machines that we want touse initially for load generation. Our initial
1. The Aurora Project was featured as the lead article in the fall 2002 issue of con-duit!, V11, N2.
SMPServer
ClusterServer
NetworkModel
ternet Lab Configuration
sed for studying a typical web service by using one of t and then using either a cluster server or multiple SMP
host the application
condu
model of this cluster is 20 dual-processor 64-bit machines. We want this cluster to beupgradable to a 40-node cluster during thesecond year.
• Networking Gear. We need sufficient numberof gigabit switches to connect the abovemachines in a local network.
• Network Storage. The lab will need an addi-tional 5Tb of disk space that would be accessi-ble by the above servers.
• Software. We would like to obtain appropriatecommercial Internet application and e-com-merce software so that we can expose our stu-
dents to the tools and
it! 2
techniques that are used tobuild real web applications.This would include suitessuch as SunOne or Web-Sphere, environments forbuilding Internet applica-tions, database systems,load-generation tools, etc.
This equipment will be used toset up the initial laboratory.During the second year, we planto expand the laboratory byobtaining the following:
• Cluster Upgrade. The smallcluster obtained during thefirst year should be upgradedto 40 dual-processor mach-ines during the second year.These machines shouldmatch the ones in the smallcluster obtained during thefirst year.
• Cluster Server. We want athird rack-based cluster
server with an additional 40
dual processor machines. This will give us atotal of 120 machines that can be used forexperiments involving peer-to-peer or gridcomputing applications. This cluster shouldconsist of dual-processor, 64-bit machineswith at least 4G of memory per unit.• Network Upgrade. We will need additionalswitches for the above machines. In addition,we will want to provide appropriate hardwareand/or software for simulating the networkperformance (packet loss, jitter, etc.) seen byan Internet application.
• Storage Upgrade. We would like to add anadditional 5Tb of disk space to the laboratoryeither by expanding the disk space obtainedduring the first year or by obtaining a seconddisk unit.
• Specialized Hardware. We are interested inexperimenting with various types of special-ized hardware including (but not limited to)firewalls, encryption, load balancers, andXML processors.
CollaborationToday it is industry that is developing Internet-scale applications and discovering their manyinherent characteristics and difficulties. We wantto have access to those problems so that we candevelop appropriate solutions and approachesthat will scale into the future. We also want toeducate students, both undergraduate and gradu-ate, to understand the problems and approachesso that they can work effectively with tomor-row’s systems. We have sent out a call for industrial partners towork with us to identify problems, help direct ourlaboratory’s configuration and uses, and provideus with discounted or donated hardware and soft-ware. We are interested in partnerships in whichboth parties are committed to working together toaddress these important and necessary Internet-related issues.The Internet laboratory is being designed as acollaborative effort. Some possible examples ofwhat we would like to see industry provideinclude:
• Speakers who would come and present tech-nical talks on Internet-related software issues.
• Visiting researchers who would spend time atBrown using the laboratory and working withthe faculty and students.
condu
Visitors to the department over the summeron our walls of a very impressive group of
Preparataworld enaof the grtime Browdents.
Franco anhuge collremarkedtaste, I’monly arrathe artwohelp frommeans ahas madmore inte
Professor Feldman came to the departmehonor to thank him for his generosity. In thissor Feldman’s left and Franco Preparata tthem is entitled “Homage to Little Bird” and w
• Sharing current and future problems involvingInternet applications. We are interested in get-ting a better understanding of what problemsthe computer industry and their customersneed to have solved in the next year, fiveyears, and ten years with respect to Internet-related software.
• Access to research tools for understandingand exploring Internet applications.
• Additional support for specific Internet-related research projects at Brown. Thiswould include both financial support and col-laborations between Brown researchers andindustrial researchers.
• The opportunity for researchers at Brown,either faculty or students, to visit and possiblywork with researchers at industrial locations.
In addition, we are looking for ways that Browncan contribute to such collaborations that are ofinterest to industry. Possible examples hereinclude:
• Providing students, either graduates orinterns, who have a detailed knowledge ofspecific Internet-related hardware and soft-ware.
• Providing speakers to present technical talkson Internet-related software issues.
• Providing access to the Internet laboratory toindustry or its customers for research pur-poses.
it! 3
will have noticed the gradual flowering art works. It’s the brainchild of Franco, whose connections with Brown’s artbled us to receive on loan a collection
aphic works of Walter Feldman, long-n Professor of Visual Arts, and his stu-
d Eli Upfal selected these prints from aection in Prof. Feldman’s office; as Eli, “The pictures you like are Franco’s responsible for the rest.” Franco notnged this loan but also shepherded allrks through the framing process (with his assistant Fran Palazzo), by no
trivial task. This handsome collectione our institutional greige walls muchresting and aesthetically pleasing.
nt on October 17 for a reception in his image, Chairman Eli Upfal is to Profes-o his right. The handsome print behind
as created by Feldman for his mother.
ACL
Mark Johnson
• Providing access to the Internet laboratory forcustomer or in-house training.
• Providing specific services using the Internetlaboratory such as beta-testing software.
• Providing independent feedback on the qual-ity or performance of software or hardwareconfigurations.
condu
AFTER-DINNERSPEECH
Space for the new Internet Lab will becomeavailable this coming January. We are currentlyactively looking for potential industrial partnersand collaborators, and are in the process of speci-fying the actual equipment for the lab. We expectthe lab to be used both for courses and forresearch in the spring semester.
The Association for Computational Linguistics(ACL) is the primary professional association fornatural-language understanding and processing. Itis a tradition at the banquet of the annual generalmeeting that the Association’s President entertainsthe members with a humorous after-dinnerspeech. The President of the Association is MarkJohnson, Professor of Cognitive and LinguisticSciences and Computer Science at Brown, andhe has agreed to let us print below his after-dinnerspeech at the Association’s banquet last July inSapporo, Japan.Friends, colleagues,As many of you know, the ACL president uses thedinner speech at the banquet to pass on insightsabout the ACL and our field in a light-hearted andamusing way. Traditionally what happens is thatthe speaker tries to say something entertaining butwinds up being facetious, embarrassing himself orherself and offending half the audience. Well, I’vebeen told that Japan is a country in which tradi-tions are very important, so I’ll do my best to keepthis one up!Instead of restricting my attention to just the ACLor even computational linguistics, I’m going totalk about something that affects virtually every-one on the planet who’s not living in a cave, andpoint out an opportunity we, as computational lin-
guists, have to do something to really make a dif-ference.I speak of course about the deep crisis that afflictsus all: a crisis of ethics. Now you may not beaware of any ethical crisis, but just open any news-paper and you'll see that scarcely a day goes bywithout a new scandal or cover-up in all levels andspheres of society: corporate, government, sci-ence, and most distressingly, even home decora-tion!In the corporate arena, consider the scandals facedby Arthur Andersen and Enron, to name just acouple. Whom can you trust when the people paidto check the books are cooking them too? Whatkind of world are we living in when corporate ex-ecutives would put at risk the state of California’selectric power just to make a few measly bucks?Even the honored fields of science and engineer-ing are not above suspicion. The funding agenciesare trying to force us researchers and our studentsto become more ethical in how we do research andtrain our students. The U.S. National ScienceFoundation requires classes in ethics for all grad-uate students in NSF training programs. And theNational Institutes of Health have gone one stepfurther—recognizing that the rot goes deeper, theyrequire that not just the students but we faculty aswell pass an ethics tests in order to get NIH grantmoney.But what really shook me up was learning that theethical rot had spread as far as that paragon ofgood taste, that epitome of thoughtful planningand organization, Martha Stewart. I’m sure mostof you are familiar with the wonderful contribu-tions Martha has made to the world of interior dec-oration, party planning and home cooking. It’shard to think of anyone who has done more toeliminate clashing patterns and colors in table set-tings, all the while encouraging innovative yettasteful dinner menus that can literally be throwntogether by anyone with a month or two to spare.So I’m sure you share my sense of dismay whenshe was accused of insider trading on the stock ex-change, selling ImClone shares on the basis of anillegal stock tip. I mean, what sort of world are weliving in if the person you rely on for advice foryour intimate dinner-party etiquette is at the sametime cheating you on stock trades?What the general membership may not know isthat the ACL executive committee has been con-sidering appointing Martha to our committee to
it! 4
During the 03/04 academic year, the CS Department is celebratingits 25th anniversary with much ado—a Distinguished Lecture Se-ries and an Anniversary Symposium and celebration banquet,among other events. Details are in a special 25th website at http://cs.brown.edu/events/25th-anniversary/ New and fascinating infor-mation will continue to be added, so do stay in touch and visit thiswebsite often. A set of faculty research summaries is on the 25th site linked to re-search already on line. The site includes a history of the depart-ment as well as amusing old photographs of faculty and staff totake you back to your student days. A live webcam has been set upin the 4th floor atrium and we already have a bulletin board on thewebsite and a connection to Brown’s alumni folks for personal dataupdates. We hope you will participate in our celebration and join us for thebig symposium and banquet next May 27!
help improve the appearance of our publicationsand generally jazz up our Proceedings. Yes, we’rethinking of making Martha the ACL Chief CoverDesign, Size and Color Coordinator.You know the problem. Members who have beento more than one or two of our conferences windup with a stack of Proceedings on their book-shelves. Up to now the size, shape and color ofProceedings’ covers have been chosen by the lo-cal conference organizers, which leads to the con-fusing cacophony of sizes, colors and evennumbers of volumes on our bookshelves. Unlikethe Proceedings of our competitor organizations,which have a pleasing harmony and color schemethat you can recognize at a glance, there’s so muchvariation in ACL Proceedings that you can’t evenalways identify them as ACL Proceedings. TheProceedings of the COLING conference in Nanteswere photo-reduced to virtually paperback-booksize. The print was so small that we should havedistributed magnifying glasses with it, like theOED.I imagine Martha’s first suggestion will be to stan-dardize the size of Proceedings; she’ll probablytell us to stick with Proceedings that are uniformlylarge. She will probably suggest that larger Pro-ceedings convey the sense of solidity and reliabil-ity we want for our field, as well as being too bigfor your colleagues to sneak off with in their jack-et pockets (Martha’s practical about this sort ofthing). I expect she will think it’s good for us to al-ternate between US- and metric-sized paper in ourProceedings to reflect the richness and diversity ofour field and make it easier for our color-blindmembers to figure out where one year’s Proceed-ings end and the next one begins.But the place I think Martha will be able to help usmost is with a color scheme for our Proceedings’covers. Her first suggestion will almost certainlybe that we switch to pastel colors, of course, butmy guess is that she’ll suggest that we try to pick
condu
colors that reflect the content of the Proceedingswhile harmonizing with the colors of our otherProceedings up there on the shelf. For example,theoretically oriented conferences would be cool,clear, ice blue, workshops with hot-breaking newresults on the latest new technology could be fire-engine red (created with a special dye designed tofade rapidly over time), and those surprise-lan-guage conferences sponsored by security agenciesthat dare not speak their name would be in invisi-ble ink on transparent clear plastic paper. Our gen-eral conferences, of course, would continue to bea generic neutral beige.I’m sure you’re as excited by these possibilitiesfor innovative new cover design as I am, so it iswith a sad heart that I have to report that Martha isunlikely to be able to actually take up her appoint-ment on the ACL executive committee; I’ve heardthat she’s devoting all of her time to her legal de-fense. I guess I should have seen this coming: itshould have been a dead giveaway when the ACLexecutive committee’s emails to her were returnednot with suggestions about doilies on the frontcovers, but with spam offering us ImClone sharesat a steep discount. I only hope that not too manymembers of the ACL executive committee spenttheir hard-earned money buying these shares—you really could have lost your shirt on these!The ethical plight of Martha and others like herwho have suffered from temporary ethical lapseshas affected me deeply. I spent many a sleeplessnight wondering if there was something we com-putational linguists could do to help her and themillions like her avoid the ethical pitfalls thatabound in modern-day life.Now the traditional response to these ethical cri-ses has been to pass laws and regulations and gen-erally exhort people to behave ethically. Forexample, the U.S. Congress has passed a numberof laws to reform the financing of political cam-paigns, with the general goal of ensuring that in
it! 5
A lowthe A
co
the future politicians will not be influenced bymoney. Well, all I can say is that you might as wellpass a law forcing rivers to run uphill. Some ofyou will no doubt point out that none of this isnews: people have been lamenting a drop in mor-als and ethics for hundreds of years, and none ofthe laws or regulations have really had any effect.Well, I would say, after two hundred years of fall-ing ethics, how low must we be by now! And isn’tit time that we did something about it?Now I’m sure that the more practical among youmust be thinking, as I am: we’ve tried all the hu-man-oriented systemic solutions, and they’ve allfailed; surely there must be a technological quickfix for this problem? This is where I think wecomputational linguists can make a contribution.Notice that most unethical behavior involves lan-guage in some way: a letter to the accountants ask-ing them to deduct a holiday home as a businessexpense, an email from an Arthur Andersen con-sultant suggesting a corporate tax dodge in theCaribbean or, in Martha’s case, a phone call froma stock broker suggesting she dump her ImCloneshares before tomorrow morning’s company newsreport.
condu
-tech solution to a high-tech problem—when C failed during a stretch of particularly hot,
humid weather this summer, fans weremmandeered and pressed into service to
prevent a machine room meltdown
This is our great opportunity. We’ve been lookingfor a killer app for computational linguistics, andI think this is it!Computational ethics would be about detecting,classifying and ultimately correcting unethical be-havior. Think how useful a little “ethics checkerand corrector” program integrated into a wordprocessor could be! It might work something likethis. The executive at Enron who first dreamt upthe idea of scalping electricity in California prob-ably sent a memo to the boss that went somethinglike this: “By artificially restricting electric-gener-ating capacity during peak summer hours we candrive electricity prices sky-high and make a for-tune”. Well, the ethics checker program wouldflag this, perhaps with a cute little icon of a clo-ven-hoofed chap with a pitchfork, and bring up awindow saying “This passage contains one ormore unethical statements. Suggested replace-ment: ‘By building more electric-generating ca-pacity in developing countries, we can raise livingstandards in the third world’”.Well, maybe the design needs a little fine tuning,and I agree some people might wonder about theappropriateness of your word-processor dispens-ing ethical advice, but I think you get the picture.I envisage ethics agents in spreadsheet programsto catch shady accounting tricks, and in email sys-tems to detect and reroute unethical messages. Forexample, someone sending me a business plan in-volving siphoning money from a certain Nigerianpetroleum-exporting corporation might receive anautomatic reply along the following lines: “DearMs. Mbuto Seke-seke, I am Mr. Johnson’s elec-tronic ethics agent. Mr. Johnson does not engagein this kind of unethical business transaction. Ihave detected four illegal business practices inyour message, and have forwarded it to the StateAttorney General for prosecution. Have a niceday!”Computational ethics seems to have great possi-bilities, with something in it for all of us. Dialogmodeling and speaker intention tracking is obvi-ously involved in distinguishing ethical and un-ethical behavior in conversations. Speech re-searchers might focus on the phonetic and prosod-ic correlates of lying, detecting those subtle pitchchanges indicating that the speaker is trying to putone over on you. There are plenty of speech appli-cations too. Voice-enabled ethical agents have ap-plications in mobile cellular phones—that, say,cut off a conversation by simulating a radio-deadzone whenever the speaker is about to say some-thing ethically dubious or compromised. There areeven theoretically interesting philosophical ques-tions: ontologists, for example, could write paperson the nature and number of different types of un-ethical behavior, and how they should be incorpo-rated into the Wordnet hierarchy.But perhaps the most innovative applications ofcomputational ethics involve wearable, pervasive
it! 6
each
e bebleo
computing. Wearable computers have been talkedabout for some time; I’m sure you’ve all heardabout “smart” fibers and fabrics that weave acomputer into your shirt or even your pyjamas.More recently some of us have been talking aboutpervasive computing, whereby a computer stickswith you all the time and records every word youutter during your life. Well, it seems to me thatwhile the technology is there or will be in a fewyears, what they lack is motivation for actuallydoing any of these things. I’m not so sure it’s agood idea to have a record of every inane thingI’ve said over the years. For example, I knew thatthe comment I made on that panel last year was ri-diculous within ten seconds of saying it; the lastthing I need is a computer continually remindingme of that fact.Now ethics checking and correction is an ideal ap-plication for wearable pervasive computers.Imagine a device that could give you a subtlewarning that you’re getting into ethically dubiousterritory and thatyou should shutup before you doanything to in-criminate your-self—how muchpain and anguishthat would save!So I’m sorry to re-port that the petindustry hasscooped us hereand we’ll be play-ing catchup for thenext few years.I’m thinking of devices like the Invisible Fence: aradio-controlled dog collar that automatically ad-ministers a series of increasingly painful butharmless electric shocks whenever the dog tries tocross a preset physical boundary. Clearly this isthe way to build ethics correctors! We can designwearable ethics correctors along the same lines asthe Invisible Fence, although in this case theboundary would be ethical rather physical. In anycase, inspired by this and other electronic pet re-straints, I envisage a whole range of ethically en-abled chokers, chalices and other high-fashionaccessories.Of course, I’m sure there are some spoilers amongyou who are going to point out that ethics check-ing and correction is a hard problem, probably AI-complete, and that we really don’t have the faint-est idea how to detect, let alone correct, unethicalactions and statements.Well, yes, but since when has that stopped us? Allthis shows is the necessity of large amounts offunding from the appropriate government agen-cies. And is the situation really any different fromthat in machine-translation, question-answeringor dialog systems? We’ve been promising to solve
is the situation rfrom that in ma
question-answsystems? We’ve
solve these profive years for alm
condu
these problems in the next five years for almosthalf a century. In fact, the one thing I can see thatdistinguishes computational ethics from these es-tablished areas of computational linguistics is thatif we ever did get it to work, it might well classifyas unethical the very research proposals we usedto get funding!And of course it’s possible that actual understand-ing really isn’t necessary for computational ethicsat all. I’m not sure exactly what empiricist ethicswould be—would it study the kinds of ethics thatpeople actually have, or, in the context of ethicsinstruction, would it give a sinister new meaningto the term “supervised training”? But clearly em-pirical ethics is a topic whose time has come.When I was planning this speech, I really didn’twant to give a talk with no solid data in it, but ac-tually writing code or doing any experimentsseemed too much like hard work. So I did whatany good empiricist would do—I used Google.The question here is: can we use corpora to auto-
matically identi-fy unethical be-havior? For ex-ample, can wesomehow discov-er that it is uneth-ical to cheat onyour income tax-es by using cor-pus data alone?Who needs intel-ligence whenyou’ve got a ter-abyte or two ofdata?
Well, Google reports 2,360 hits for the query “eth-ical cheat income taxes”, but only 795 for the que-ry “unethical cheat income taxes”, which gives anodds of 4:1 in favor of cheating on income taxes.(The nitpickers amongst you will be pleased toknow that all the statistics reported in this paperare significant at p less than 10-4!). Interestingly,you get the same sort of results for the query“scalp electricity in California” (odds of 8:1 in fa-vor), and “dodge tax in Bahamas” (5:1 odds in fa-vor). The actual hit counts go up if you replace thewords “ethical” and “unethical” in the query withthe words “good” and “bad”, but otherwise the re-sults stay pretty much the same.What does this mean? While I don’t expect Webcitizens to be perfect, I had hoped that on averagethey would at least lean the right way! But per-haps the ethics of dodging taxes in Bermuda andmaking super-profits by reselling electricity inCalifornia is a little subtle for the highly caffein-ated hackers who put the Web together, so I decid-ed to look at a couple of ethically less complicatedcases. I’m disappointed to report that even “beat-ing your children” and “murder your in-laws”come out with odds of 5.5 in favor!
lly any different ine-translation, ring or dialogen promising to ms in the next
st half a century
it! 7
THE 31MOBI
Host UgurCetintemel
My goodness, what sort of world are we living in!Do 5.5 times more people really think it is OK towallop the kids and knock off granny if they hap-pen to get in the way? And more importantly forus science types, is there hope that we can get thisto work well enough for a research grant propos-al? Is empiricist ethics doomed to track the ever-lowering ethical standards of society at large? Oris there some kind of hack we can use to get thekind of normative judgments we know the goody-goodies in the funding agencies will require? In-spired by the immortal words of Benjamin Disrae-li and Mark Twain about “lies, damned lies andstatistics”, I pressed on.The first thing to note is that the Web seems to bepopulated by cheery, perpetually upbeat people —silly fools—who use the word “ethical” some3,700,000 times, which is 3,100,000 times moreoften than they use “unethical”. The same sort ofthing holds for “good” and “bad”. Well, clearlythe Web is unfairly biased towards nice, happy,ethical terms and doesn’t reflect how bad thingsreally are. Can we correct for this unrealisticallypositive view of the world somehow?Now bear with me through a little statistical mum-bojumbo so we can get the sort of results we want.We can correct for the over-optimistic bias by ask-ing not just whether dodging taxes is ethical (ac-cording to Google, the majority of documents
condu
ST IPP SYMPOSIUMLE & PERVASIVECOMPUTING
seem to suggest it is), but asking instead whetherdodging taxes is any less ethical than the myriadof other ethically dubious things that people onthe Web talk about. And when we do this, viola!we get the results we want! I won’t bore you withthe details, but if you calculate the log odds ratiosfor all of the searches I mentioned earlier they allcome out negative, which is appropriate sincenegative values set the right sort of moralistic tonewhen we’re talking about corporate tax havens,scalping electricity and beating children. Even“murdering your in-laws” turns out to be ethicallynegative, which is just as well for everyone in theaudience with married children.I certainly don’t offer this as the final word in em-pirical ethics, but just to show the kinds of possi-bilities there are out there and the kinds ofmischief you can get up to with Google and a bot-tle of red wine. So, I propose to you that next timeyou’re looking for a project, consider computa-tional ethics. We can do this in five years! Thinkof how it would help the world; the right little pro-gram could have pointed out to the Enron execu-tives that scalping energy in California wasunethical, and it should have warned Martha tostick to interior decorating. The time has come toask not what Martha can do for us, but what wecan do for Martha. And if anyone is interested ina very special deal on some ImClone shares, seeme after the talk. Thank you and good night.
Recent years have witnessed astrong trend towards mobile andpervasive computing, in whichlarge numbers of casually acces-sible, mobile or embedded com-puting devices are connected byan increasingly ubiquitous net-working infrastructure. The con-fluence of increasingly small andpowerful computing devices andubiquitous wireless and wiredconnectivity is radically chang-ing the way we live, work, andplay. Middleware systems andtechnologies that make possibleand drive mobile and pervasivecomputing were the topic of our31st IPP symposium on May 6,
2003. The symposium brought together speakersfrom Fidelity, Hewlett-Packard, IBM, Intel, In-vensys/Foxboro, and Sun Microsystems.
I inaugurated the symposium with an overviewtalk that attempted to lay out the groundwork forthe rest of the day by focusing on the past, present,and anticipated future advances in hardware andsoftware technologies that make possible mobileand pervasive computing. I also talked about relat-ed research projects we are currently undertakingin the department. Up next was our first guest speaker: Archan Mis-ra of IBM Watson. Archan talked about recentresearch in distributed middleware for data com-position from pervasive data sources. Archanidentified decentralization, scalability, and adap-tivity as the key design considerations for next-generation distributed middleware. He describedwork by him and his colleagues on Context-Sphere/CLASH, an adaptive middleware infra-structure that enables a large base of servers to beused as a shared computing resource for support-ing large-scale pervasive services and applica-tions.Phillip Gibbons of Intel Research addressed ap-plications involving a large number of Internet-connected rich sensor units (such as web camerasand microphones). Phillip presented the IrisNetsystem, which strives to provide a generic infra-structure to ease the development of sensor-en-riched Internet services. IrisNet facilitates
it! 8
Symposium speakers, clockwise from top left:Ugur Cetintemel, Brown; Archan Misra, IBM Watson; Phillip Gibbons, Intel
Research; Hesh Kagan, Invensys/Foxboro; Aad van Moorsel, Hewlett-Packard Labs; Brad Hampson, Fidelity Investments; Jim Waldo, Sun Microsystems
scalable distributed query processing at the gran-ularity of individual sensor units, an ability com-monly demanded by many sensor-driven ap-plications. Phillip also showed us a videotapeddemo of the IrisNet prototype running a mockparking-space finder service, which would nodoubt be extremely popular here on the Browncampus.The morning session was concluded by Aad VanMoorsel of Hewlett-Packard, who discussed thescale and complexity challenges in developingcomputational utility grids for pervasive applica-tions. The goal is to make computational resourc-es as easily accessible and usable as traditionalutilities such as water and electricity (e.g., “I canjust plug my CPU-free terminal in the CPU jackand run my Mathematica, regardless of where Iam”). Aad finished his talk by arguing the neces-sity of using standardized web-services-basedmiddleware in order to ease the job of IT adminis-trators and to make pervasive applications moremanageable. Jim Waldo of Sun Microsystems started the af-ternoon session with an engrossing and amusingtalk that identified some key middleware chal-lenges in pervasive computing. Jim argued thatnext-generation pervasive applications would re-quire a level of reliability and adaptability thatcurrent software systems cannot achieve. Accord-ing to Jim, it makes more sense to try to build sys-tems that can deal with failures than to try to buildparts that don’t fail. Jim also made the case forsystems that can evolve continually as changes
condu
happen (“change the fan belt while the car is run-ning”), no matter how drastic the changes are(“change the engine while the car is running”).The last two talks addressed the impact of perva-sive computing on two different applications do-mains. Brad Hampson of Fidelity Investmentsspoke about “Natural Broker”, a voice-driven que-ry interface to a financial information system thatprovides basic statistics about publicly tradedstocks. The motivation for developing NaturalBroker was to make wireless PDA-based financialapplications easier to use. Brad talked about thetechnical problems his group had faced while de-veloping Natural Broker and concluded his enter-taining talk by pointing out that customer demandfor wireless query interfaces was still unclear.Finally, Hesh Kagan of Invensys/Foxboro char-acterized the effects of pervasive computing on in-dustrial process automation. Invensys/Foxboro,one of the leaders here, is very much concernedwith providing real-time monitoring and controlof their systems. After an eye-opening introduc-tion to the world of process automation, where ev-idently even the small process lines have hugenumbers of embedded control units, Hesh dis-cussed future plans to use wireless sensor net-works for non-obtrusive monitoring of processlines.
it! 9
COMMUNICATIONS FROM ALUMNI
Gideon Mann (l) and Brock Pytlik
DON BLAHETA, PhD ’04Quoth Fran Palazzo: “Hope you are doing well,Suzi is asking me for info about recent grads.Please let me know what you are doing. Thanks.”I started writing a simple response, and it sort ofmorphed into a conduit-style chatty update. Notsure if that’s what you were looking for, and if not,feel free to edit it accordingly, but I figured Imight as well leave it as is. :)What am I doing? Enjoying myself! I’m an As-sistant Professor of Computer Science at KnoxCollege in Galesburg, IL. This term I’m justteaching intro; it’s in Java, but I’ve been recyclingsome of the stuff I helped write for CS17, lo thesemany years ago. (Just today I handed out the thirdproject—Eliza!) Next term I’ll teach another sec-tion of intro, plus an NLP class, and beyond that,who knows what the future holds? I’m one of justthree CS profs here, so I’ll get a chance to teach alot of different classes. Not much research so far,but as I settle in I’m finding more time for thingsother than class prep, so hopefully soon I’ll beable to get back to that.Those who followed my extracurricular activitiesat Brown will be totally unsurprised to hear thatI’m trying to get a ballroom dance club goinghere. I’ve been teaching a class every Sunday andI have a small group of regulars (both fac/staff andstudents) that, hopefully, will grow as people telltheir friends. Only time will tell. Outside Knox, I think I’m becoming knownaround town primarily as “that guy who drives thecute car” — I just bought a blue and white MiniCooper, the only one in Galesburg, and quite a fewpeople have been commenting on it.Anyway, I’d love to hear from other Brownfolks; my email here is [email protected].
ADI GANZ, ScB/ScM ’03As a member of the University’s top-level ad-ministrative staff (VP Research), Andy vanDam received the following email messagefrom Russell Carey, VP and Secretary of theUniversity:Each year, as many of you are aware, we askone undergraduate student in each degree cat-egory to be the symbolic degree recipient inthe ceremony on the Main Green. Adi Ganz isthe sole recipient this year of the combinedScB/ScM degree. Quite unexpectedly, in hisreply to this request he wrote a paragraph thatI thought captured quite nicely what we hope
condui
the undergraduate experience will be for students.I thought his perspective on his Brown educationmight be of interest to you, so I am forwarding it,with his permission.Dear Mr. Carey,It is an honor to accept your offer.Brown has left a positive permanent impact on mypersonality and academics. It allowed me not onlythe gratifying opportunity to pursue a combinedScB/ScM degree, but also the opportunity to helpfellow students through various channels, mostnotably my six teaching assistantships in Comput-er Science and Biology, and to incorporate educa-tion in the life sciences into my computer sciencecurriculum. This allowed me to pursue my dreamof applying to MD/PhD programs and become amedical scientist. If there is any way in which Ican help you or the President to further strengthenthe message that students have a vast potential formultifaceted growth at Brown, please let meknow.Sincerely, Adi
GIDEON MANN, ’99 andBROCK PYTLIK, ’03This picture of Brock and me, Gideon Mann, wastaken in our office at Johns Hopkins. We are bothPh.D. students of David Yarowsky in the Centerfor Speech and Language Processing. WhenBrock came this fall, he reminded me how excel-lent the Brown Computer Science Department isand how much I enjoyed being there. Please givemy fond regards to the faculty, staff and all of theunredeemable goofballs from ’[email protected]
t! 10
GAIL MITCHELL, PhD ’93Hi Eugene: I’ve been meaning to send a note toyou ever since reading your ‘unplugged’ article[about the creator of the smiley :-)] in the springissue of conduit! to let you know that I don’tknow Heidi Fox, but I’m currently managing aproject for which Ray Tomlinson (the @-signguy) is the Chief Architect. His office is threedoors away (a corner, of course. I at least have awindow). I often notice news photographer typesaround interviewing from all over the world. Iguess his recent IEEE Internet Award prompted atleast one of those.When I brought up the conduit! citation ;-) in ourweekly management meeting, the PI of the sameproject, Bruce Robert, mentioned that he knowsyou (if I’m remembering correctly).Hope all is well with you. I guess you may haveseen my youngest wandering around the officethis week :-). He much prefers Brown to BBN asa place to spend a day.Gail
SCOTT RAPOSA, ScB ’94 About two years ago I had just left my position ata dot-com called iHarvest in Silicon Valley. I hadlived in California for almost four years and reallygotten into the lifestyle, but something was tellingme I needed to get away from it all—from thehigh expenses, from the hi-tech atmosphere, fromthe corporate world, from everything I’d grownaccustomed to. So I did what any sane personwould do—I quit my job, sold just about every-thing I owned (including my beloved Miata whichI still miss dearly), packed all my stuff into afriend’s SUV, and moved to the backwoods ofsouthern Oregon. However, my Oregon experience ended. Quickly.I decided in about three weeks that I had to getaway, far away: the specific cir-cumstances combined with a housefull of leaks, dogs, rats, bats, andseveral unnamed woodland crea-tures, weeks of ceaseless rain and40-degree weather made my deci-sion clear. Now, not having a car, ajob, much money, or any clue what-soever about anything really, I de-cided to camp out with my brotherfor a month. And why not, seeingthat he was in Florence, Italy?Needless to say, my stay in Italymore than made up for my “dis-comfort” in Oregon. There aresome great (and somewhat bizarre)pictures of the trip here on mybrother’s site: http://www.robertraposa.com/gior-nale/gennaio_02.htm
condui
http://www.robertraposa.com/giornale/febbraio_02.htmMy next destination was my a-bit-too-rural-for-my-own-tastes-but-hey-you-gotta-take-what-you-can-get hometown of Rockaway, NJ. I’ll say thisfor Rockaway: the deer are real friendly. I evenmanaged to pet one. And my parents are greathosts and great cooks, so you won’t catch me com-plaining. The best part of my stay in Rockawaywas that I could afford to stay unemployed for 18soul-searching months. But while yes, it was greatto have 18 months away from the grind, it’s notsomething I’d recommend to the faint of heart. Forme, at least, not having a job meant not having apurpose or an identity and therefore having tocome face-to-face with all kinds of things I didn’twant to know about. Fortunately, I found my way out of the depressionby discarding the job listings for all the jobs Ididn’t want in the first place and focusing on myonly source of inspiration at the time: my key-board (music workstation/synthesizer/expensive-toy-that-makes-cool-music). (Actually, buyingthat keyboard, a Korg Karma, was one of the rea-sons I didn’t have much money left.) So I decidedto use my investment wisely by learning it insideand out. I began composing on it and program-ming it and became active on an online forum ded-icated to the keyboard. After a few months, Idecided I would sell my work—a set of newsounds designed specifically for the Karma key-board. Shortly thereafter, I contacted the guy whodeveloped the technology that runs the keyboardto see if he wanted to collaborate. Fast forward a few months: my product, calledReincarnation (get it? Karma? Reincarnation? Ohwell, I liked it) was co-produced and marketed byStephen Kay, the inventor of the Karma technolo-gy, and I began working for his company, KarmaLab (which is now just the two of us) to developthe core music technology further. Reincarnation
t! 11
Mark with his 1979 Piper Archer, the “Kitty Hawk,” enjoying fall foliage in New
Hampshire this October
goes for $50 and is selling well, andthe Reincarnation web page (http://www.karma-lab.com/sounds/rein1.html) has some nice user re-views. Unfortunately, there aren’tany online demos to listen to (yet),and you can’t do anything with myproduct unless you have a $1500keyboard to go with it; but you canstill take a peek if you’re interested.So these days I spend four days aweek working from my studio inmy new home in Montclair, NJ. Ishare two floors of a huge Victorianhouse with my housemate Kris andher two cats. We each have two bed-rooms to ourselves, our own bath-room, and our own living room. Weshare the kitchen and a small “meditation room”(yes, she’s into all that stuff too). I’ve been per-forming on my keyboard and plan to do a lot moreof that soon. For about eight months, I played in asmall band every Sunday morning at a spiritualcenter. It wasn’t your average Sunday service mu-sic as this group likes to get up and dance! I alsoplayed background music during the meditationwhich taught me a lot about how to create a musi-cal atmosphere without distracting the listener.Now, I’m starting to play more chill-house, ambi-ent groove stuff at cafes and parties and such, adream of mine for the last few years. I’ll end by saying that I’m living a life now that Inever, ever, ever in a million years could haveimagined just a short time ago. For one thing, Ilive in NJ! (OK: it’s really not that bad after all. Ifought it for a long time, but Montclair is cool andI’m only 12 miles from NYC. I truly miss themountains, but hey, easy access to real pizza andreal bagels nearly makes up for the loss.) But seri-ously, at least a few times a week I experiencesomething amazing—like finding an incredibleplace to live in just one phone call, or being of-fered an antique bedroom set worth thousands ofdollars for only $100. Call them miracles, callthem whatever, but I know they aren’t about luck.My expectations for what I’ll get out of life are ex-tremely high, and getting higher each day. Andthese expectations are being reflected in my expe-rience. It’s been quite a ride to get here, but I’vebeen blessed with some amazing spiritual teachers(who knew they were on the east coast too?!) andfriends. Bottom line: things are good.With much peace, joy, happiness, love, and allthings good. [email protected]
MARK STERN, ScM ’91(Steve Reiss’s wife Loretta, ScM ’81, who becameinterested in issues concerning communicationwith deaf people in the course of her doctoral
condui
work, knew Mark Stern while he was at Brown andrecently noticed his name in an article about deafpilots in the August/September issue of NADmaga publication of the National Association of theDeaf. She passed it along to Suzi Howe, who madecontact with Mark.)I’ve enjoyed reading conduit! over the last severalyears, so here’s my turn. After getting my ScM inComputer Science at Brown, I worked in theMacintosh Human Interface Group at Apple forfive years. I joined Netscape in 1996 as a seniordesigner for their browser, and then managed de-sign teams for Netscape’s server products andNetscape.com. I continued with the company afterthe merger with AOL as Director of Product De-sign for AOL’s Web Properties. During theseyears, I pursued my dream of learning to fly, ac-quired a small plane, and in 1999 flew it throughall the lower 48 States. You can read my chroni-cles of this aeronautical journey at www.flight48.com. Last year, after 20 years of living and work-ing in Silicon Valley, I “retired” from the comput-er industry and took some time off to move backto the East Coast and to hike in New Zealand.Now, in a new career, I am fixing up old houses inthe upper Connecticut river valley (near Dart-mouth).Mark Stern, White River Junction, VT.
DILIP D’SOUZA, ScM ’84In the fall 2002 issue of conduit!, Dilip’s book,The Narmada Dammed: An Inquiry into thePolitics of Development was highlighted in the‘changelog’ column. He wrote to let us knowabout a PBS production about the Narmada dam.Much of what’s in this documentary is what Iwrote my second book, The Narmada Dammed,about. It was shot in many of the same places thatI travelled in, some at more or less the same time.
t! 12
IVICO
User Interface/3D Graphic Research
Scientist Andy Forsberg
I saw a slightly different version of it here a coupleof weeks ago, and Franny Armstrong was there todiscuss it with the audience. Worth seeing if youget a chance: a good intro to the [email protected] Howe forwarded Dilip’s email to the facultyand John Hughes was moved to reflect upon asimilar situation:
One of my kids’ favorite books a few years agowas about the building of the central-MA Quabbinreservoir that provides water for Boston. It’scalled Letting Swift River Go. It ends with theprotagonist, now a grown woman but a child at thetime of the damming, out rowing on the reservoirwith her Dad on a summer evening. He’s pointingout where various things used to be, and shewrites:
I leaned over the side of the boat andcaught the starry water in my cuppedhands. For a moment I remembered the
condui
NTERACTIVEN-BASED DEMO
Jonabreak
wind through the willow, the trains whis-tling on Rabbit Run, the crossroads whereI had met Georgie Warren and NancyVaughn. Gone, all gone, under the waters.Then I heard my mother’s voice coming tome over the drowned years. “You have tolet them go, Sally Jane.” I looked downinto the darkening deep, smiled, and did.
It’s interesting that we have something in ourvery backyard that is rather similar to what Diliphas been looking at and writing about. One in-teresting lesson is that the effect of such a flood-ing lasted (in this case) only about a generationand a half. By the way, the drawings are by Barbara Cooney,who seems to me to be the best kids-book illustra-tor around, now that Robert McCloskey is dead.
David Erickson’s finger and hand move-ments are tracked as he “paints” on a wall in
Last summer, collaborators fromLos Alamos National Lab (LANL)visited the Graphics Group to expe-rience recent work done at Brownfirsthand. One demonstration was avirtual painting program in whichyou could put virtual ink on a phys-ical wall by touching the wall withyour index finger. When and whereto display ink was determined bythe 3D tracking system and soft-ware algorithms.Michael Black and Bill Warren pur-chased the Vicon camera-basedtracking system for computer vi-sion projects and action and percep-
tion experiments in their one-of-a-kind motion-capture facility. LANL recently installed thissame tracking system in their Cave and work done
for this demo will be used to exchange Cave soft-ware in the future. David Erickson (CS ScB, ’04)and Jonathan Bankard (CS/Econ. ScB, ’05)worked at Brown over the summer to supportprojects that used the Vicon system and anothercamera-based system. This “toy” painting application helped us betterunderstand both how to use the Vicon system forinteractive applications, and the accuracy and la-tency of the system.
t! 13
than Bankard and David take a from technical work to exercise
their brains’ right sides
the Moonlab. Three markers are tracked by the motion-capture system and movement is
interpreted by the painting application
UGUR CETINTEMEL. In the spring,Ugur was a panelist in the plenary panel session(entitled “Middleware for Distributed Sensor Net-works”) of the International Conference on Dis-tributed Computing Systems 2003, which washeld in Providence in May. During the summer,Ugur served on the program committees of the In-ternational Conference on Data Engineering 2003and Data Engineering for Wireless and MobileAccess 2003. Ugur was also involved in the orga-nization of the Workshop on Internet Applications2003.
TOM DEAN. Tom has been appointed deputyprovost at Brown and consequently has been su-premely busy.
condui
DAVID DURAND. David’s research activityhas been very light, as his non-Brown commit-ments have been high. He gave a talk “Trees Con-sidered Unnecessary: Event-based Regular Trans-formations for XML” at the E-Biosci/Oriel Annu-al workshop (http://www.e-biosci.org/sept/pro-gramme.htm). He has continued to work on theText Encloding Initiative (http://www.tei-c.org/)as a work group member and member of the TEIcouncil. Perhaps most exciting is that he will beteaching the Document Engineering course againin the department this spring. Said he, “It wasgreat fun last year, and should be even better thisyear.”
AMY GREENWALD. Ella GreenwaldBoyan was born at 12:50 a.m. on June 2nd at theAlternative Birthing Center in Woman and InfantsHospital. (Yes, that’s right, the ABC: i.e., withoutmeds!) She went home at 12:50 p.m. that sameday, and she and her family spent a blissful after-noon in the backyard of their new home right herein Providence.Ella’s first month was restful. But by two months,she was already meeting with Brown’s TAC team,helping them design their entrant “Botticelli” inthe International Trading Agent Competition. Shedecided against going to the competition at IJCAIin Acapulco (too many mosquitoes), and sent twoproxies instead: Mike Benisch, representing Bot-ticelli, and Jesse Funaro, representing Brown’s
Biking enthusiasts Shriram Krishnamurthiand Mark Dieterich co-organized somegreat rides on the East Bay bike path overthe summer. Students, staff and facultyparticipated in the 20-mile rides.The startof one is pictured below — Michael Blackcaught up with them later...
Said Shriram, “We took the photo at Del’s because itwas unimpeachable evidence that we had indeed rid-den at least ten miles each way!”l to r: Hulya Yalcon, Shriram, David Laidlaw, TomerMoscovich, Guillaume Marceau, Liz Marai, ThomasHofmann, Peter Sibley, Mark Dieterich and MichaelBlack.
t! 14
nduit! 15
coClassic entry, RoxyBot. Both agents were finalistsin their respective competitions.By the end of August, however, Ella was ready forconferences. She and her family flew down toWashington, D.C., where her mom gave two talks:one at ICML, on reinforcement learning and cor-related equilibrium, and a second invited talk atCOLT, on a general class of no-regret learning al-
gorithms and game-theo-retic equilibria. Betweentalks, Ella escaped to Vir-ginia to rally for HowardDean on his “SleeplessSummer Tour.” Being adoctor, Howard refused akiss from Ella, but he didagree to pose for a photo-graph.
MAURICE HERLIHY. Maurice’s paper onwait-free synchronization won the 2003 EdsgerW. Dijkstra Prize in Distributed Computing. “Her-lihy’s paper has been extremely influential inshaping the theory of distributed computing.”
JOHN HUGHES. Spike is on sabbatical inGrenoble, France— “More mountains, less sail-ing. Lots of hot weather and good cheese. Then Iturned around and came back to San Diego forSIGGRAPH a week later. But now I’m back inGrenoble, puzzling over questions like “What’sthe French word for ‘french doors’?””
SHRIRAM KRISHNAMURTHI. Shr-iram had a fun spring and summer working with avery talented group of PhD and undergraduatestudents. The two latest (under)graduates, BrockPytlik and Colin Blundell, are off to Johns Hop-kins and Penn, respectively. Shriram also graduat-ed his first PhD (co-advised) in the spring, now ata naval research lab. He served on the programcommittees for Component-Based Software En-gin-eering and Automated Software Engineering.He helped boost the economy by buying a new bi-cycle, which he’s bumped, banged and dentedwhile already putting on several hundred miles.The stuffed animal menagerie in his office has ex-panded to welcome an emu and various marinefauna. He also acquired a Napoleon hat in Mainz,just before Fastnacht.
JOHN SAVAGE. John gave an invited ad-dress “Computing with Electronic Nanotechnolo-
gies”at the 5th Italian Conference on Algorithmsand Complexity in Rome at the end of May. At theend of June he completed his service as a facultyofficer during an important time in the history ofBrown. In May the Task Force on Faculty Gover-nance, of which he was Chair, voted itself out ofexistence after completely overhauling Brown’sfaculty committee structure in about a year, themost extensive revision in faculty governancesince 1969. During the last academic year Johnalso served on the new Academic Priorities Com-mittee, introduced by the Task Force, to bring se-nior faculty members together with senioracademic administrators to set priorities and rec-ommend the allocation of resources such as the100 new faculty positions recommended by thePresident and authorized by the Corporation.
ROBERTO TAMASSIA. Roberto wasawarded a prestigious NSF Information Technol-ogy Grant for a $2M project entitled “Context-Aware Computing with Applications to PublicHealth Management” in collaboration with re-searchers at the University of Illinois at Chicago(UIC) and at Northwestern University. Thisproject is led by former postdoc Isabel Cruz, nowAssociate Professor at UIC. Roberto also receivedan NSF grant on his project “An Algorithmic Ap-proach to Cyber-Security” in collaboration withMichael Goodrich (University of California, Irv-ine) and is a co-PI on another recently awardedNSF grant, the “Brown Internet Computing Labo-ratory” project led by Steve Reiss. In the summer,Roberto started a research collaboration with SethProctor at Sun Microsystems on an informationsecurity project that involves PhD student Dan-feng Yao. Sun has provided a research gift in sup-port of this project. In September, Roberto gave aninvited plenary lecture on Authenticated DataStructures at the European Symposium on Algo-rithms in Budapest. This talk includes recent re-sults developed with his PhD student NikosTriandopoulos. Roberto’s widely adopted text-book on Data Structures and Algorithms inJava (coauthored by Michael Goodrich) is now inits third edition. A forthcoming Korean transla-
SHRIR(and
ShriramKrishnamurthi
tion is the latest of a growing list of internationaleditions.
ELI UPFAL. Eli was the organizer of RAN-DOM GRAALS 2003, (Random Graphs and Ran-domized Algorithms) in Bertinoro, Italy. He wason the program committee for the 10th Collo-quium on Structurual Information and Communi-cation Complexity (SIROCCO 2003) in Umea,Sweden. He was also on the program committeefor the Second International Workshop on Exper-imental and Efficient Algorithms, in Ascona,Switzerland.
condui
AM UNTETHEREDnearly unhinged)
PASCAL VAN HENTENRYCK. In Au-gust, Pascal traveled to a conference in Acapulcoto finish a paper with Aris Anagnostopoulos andRussell Bent, drawing some inspiration from thescenic bay and the pool (they also presented somepapers at the conference). In September, Pascaltraveled to Ireland with Cora Borradaile and Rus-sell to finish another set of papers, drawing inspi-ration from the scenic coast and long runs on hillyroads. The flight crew at Cork had a fight beforetakeoff, making the trip back that much more ex-citing.
The sender’s name was in all-upper-case, and the subjectline read, “Not exactly...”.Spam.But I’m glad I opened it any-way. “This is not exactly yournormal query...”, the messagebody began. It was from an up-standing citizen at the NASAJohnson Space Center, askingfor... the Export Control Clas-sification Number (ECCN) forDrScheme, a Scheme pro-gramming environment Ihelped develop.The what for what?!?
It transpires that a tasteful astronaut (name with-held to protect the innocent) had loadedDrScheme on his laptop for use during his freetime in space. He was (a) going to be on the Inter-national Space Station (ISS) and (b) launching ahard drive image from Russia. The ISS is a multi-national enterprise and Russia is, well, anothercountry, so the act of sending software in both in-stances involves an export. Both, consequently,need an ECCN. (I wonder if there is some tacit as-sumption here that outer space, or at least the ISS,is a foreign land. I wonder if there isn’t somegroup squirreled away trying to determine theequivalent of nautical limits in space.)The US government does a remarkably good jobof publishing documents on the Web; anyonewho’s had to deal with the IRS or the lucky fewwho’ve had to contend with the INS will grudg-ingly admit this. These are sometimes no morethan OCR-scanned copies, but the scanning yieldsenough clarity that Google can find the docu-ments. I can’t begin to imagine what this process
must have been like ten years ago. (Probably a lotsimpler: I’d have written off for an official docu-ment, then returned to my regularly scheduledwork.)Anyway, accessing prose is not the same as under-standing it. The second paragraph of the first doc-ument I read began “The CCL is contained inSupplement No. 1 to part 774 of the EAR.”It was going to be that kind of day.I have some layman’s thoughts about such prose.It doesn’t look like natural language at all. I con-jecture it’s because natural language is rife withanaphoric references. Letting context and a richlanguage of reference “do the talking” leads tohigher communication bandwidth, but it’s hell ondisjointed documents.... But this is beginning toread like one of Eugene’s articles, so I should stopthis digression now.The ECCN system is quite simple once you getthe hang of it. Numerous artifacts, from nuclear tosoftware, are given codes such as 14D993. Thesecodes determine the range of permitted distribu-tion and the licensing demands on those who wishto acquire them. Everything else is assigned thedefault code EAR99, sometimes designated NLR(No License Required). So it’s simply a matter ofreading enough documentation, slotting yourproduct, and finding a code. This is a bit like say-ing that feeding your pet octopus is simply a mat-ter of pulling aside the arms, placing the food inthe middle, then retracting your limb. Easily done,but you may lose an organ or two getting the hangof it.Reading these documents was not without its re-wards. I discovered, for instance, that EAR 740.13(d) (3) (ii) (A) differs from EAR 774 SupplementNo. 2 (2) (a), though they describe the same thing.The former document is dated later, yet leaves outa key provision included in the latter, the recogni-tion that software may be sold electronically! Bet-ter still, I found that ECCN category 14D993specifically restricts the export of “program”proof and validation “software” using mathemati-cal and analytical techniques and designed or
t! 16
CHAR
EugeneCharniak
modified for “programs” having more than500,000 “source code” instructions. Beware, allyou verification researchers—don’t get too ambi-tious!Having done my research, I determined that wefall under License Exception TSU (Technologyand Software — Unrestricted), though it all gets abit sketchy here because our software is not spe-cifically sold. (The concept of free softwaredoesn’t seem to be in the ECCN vocabulary.Cheap political crack omitted here.) This didn’tnecessarily mean we were EAR99, and EAR99 isan ECCN, whereas TSU isn’t, and NASA hadasked for an ECCN. Mustering great courage, Icalled the encouragingly named Outeach and Ed-ucational Services Division of the US CommerceDepartment. After only a few minutes, I was con-nected to a gruff counselor.“Can I help you?”“Ah, yes, I was hoping to get a quick clarificationon an ECCN classification.”(gruffer still) “Yes?”(deep breath) “If I find that my product falls underLicense Exception TSU under part 774 of theEAR, can I assume it has the ECCN of EAR99?” I believe I did this without inhaling.These are the moments that test the mettle of greatmen. Like Major Major’s father, my counselorwas made of stern stuff. He paused for just a mo-ment, just long enough to accord respect to some-one he clearly perceived to be a fellow numbers-
condui
NIAK UNPLUGGED
and-policy wonk. And then, in a voice rich incameraderie, he said, “You know, I’ve been tryingto determine that for five years myself!”Being excepted from a license was not the same asnot requiring a license? In the logic of the US gov-ernment, did the Law of the Excluded Middle notapply? What was the difference between the two?Syria.No, seriously. The difference really is Syria. TSUpermits export to Syria, but EAR99 does not.My counselor told me this with glee. Apparently Ihad hit on one of his favorite trivia questions, be-cause, now really warming to the topic, he in-formed me that he had asked this very question ofmany of his colleagues. Some had incorrectlyidentified the two; others, quicker on the draw,had said that they were indeed different but hadrecommended the use of ECCN 4D994. As every-one knows — shucks, by this point even I knew —this was a contemptible response, because 4D994permits the export of only specific software prod-ucts. And so on.Anyway, this story has a happy ending. We appearto meet EAR99. NASA Johnson has acknowl-edged receipt of our ECCN. I’ve been in touchwith the fine astronaut. DrScheme will soon behappily beta-v-cs-reducing expressions in outerspace.Where has your programming language been to-day?
In response to my piece in the last‘unplugged’ about my studentSharon Caraballo, who had herPhD rubber chicken mounted by ataxidermist (see next page), I gotthe following delightful letterfrom another of my PhD students,Jim Hendler.
Eugene - I was thrilled tosee the mention of rubberchickens in the Conduit. Iam pleased this traditioncontinues, and I don’t
know if you are aware of it, but one ofyour students, to wit, me, was the veryfirst recipient of a Brown PhD RubberChicken! The story may not be wellknown, so I share it with you here.
As you may recall, the tradition in our de-partment once upon a time was that thedefense was mainly a formality — the
presentation was given, a few simplequestions were asked, and then back to thedepartment for the party.
For my defense, however, you invitedDave Waltz, a well-known AI scientist, asan outside reader. After my presentation,Dave asked a couple of really hard ques-tions. Motivated by his example, PeterWegner and John Savage decided to showthey too could ask hard questions, andthus I spent about 45 minutes in the hotseat while questions were asked as to thevalue of the work, the novelty of the con-tribution, and other such FAQs in a PhDdefense (although not often at Brown pri-or to this).
When I got back to the department, I wasgrabbed by a few of the more interestinggraduate students and taken downstairs toone of the grad offices, where a rubberchicken had been living that year (for areason none of us can remember). Thebird was presented to me by Robert “Bob-bo” McCartney and a couple of otherswith the admonition that they had never
t! 17
den, Franco ofpictured outsidements and is en
heard such “chicken-(expletive-deleted)questions” and that I deserved this rubberchicken for putting up with it. And thuswas a tradition born...
By the way, this much-travelledrubber chicken is still proudly
displayed in my office at theUniversity of Maryland. It hasstarted to deteriorate after 17years, and I’ve been worriedabout how to preserve it —thanks for the tip about taxi-dermy!
p.s. I cc Suzi in case you allfeel this is interestingenough for a Conduit letter,
if not, at least share it withTrina and others who might
enjoy knowing where the tradi-tion originated.
p.p.s. The other souvenir of my de-fense, also still in my office, is the meatcleaver discussed in my thesis. Let’s beglad that that wasn’t the tradition that gotstarted there...Professor James Hendler, PhD ’[email protected]
condui
Franco Preparata de-signed and created thisbeautiful leaded glasscorner window piece forJennet Kirschenbaum.After seeing an old-fash-ioned iris from her gar-
fered to make a stained glass version. They’re Jennet’s office, where it attracts many compli-joyed by all passersby
I should note that in subsequent email Jim men-tioned that the PhD student after him got a rubberturkey, but could not remember who that was. Imention this because it comes up in later detectivework.
At any rate, I found this very interesting becauseit conflicts with a version that I published in ‘un-plugged’ a few years ago, to wit:
The conferring of the rubber chicken(which is thrown at, not handed to, thesuccessful candidate) goes back to 1985or so when, after a particularly bad invitedtalk, one graduate student commented toanother that someone should have shut thespeaker up by throwing a chicken or somesuch at him. It turned out that the personsuggesting this was due to defend hisPh.D. thesis a few weeks later, and severalof the graduate students thought that dur-ing the defense they would throw a chick-en, or at least a rubber chicken, at him.They chickened out (so to speak) andthrew it only after the talk, but a traditionhad started.
The dates correspond, (Jim’s CV lists his PhD asMay 1986, but he could have actually finished upin 1985) but there are a lot of differences betweenthe two versions. I asked Jim about this but got asresponse a dignified silence. I sent email to RobertMcCartney, who plays an important role in Jim’snarrative. In response I received Robert’s recol-lections, plus what must be the original documentthat I read before creating my earlier history. Writ-ten by Tom Freeman, another of our early gradstudents, it not only sides with my version but alsocontains a lot of the same phrases, which I musthave read and unconsciously parroted (e.g., “par-ticularly bad .. talk”). (Trina Avery comments that“Now you know how classicists spend their time”,and she should know, she R-ABD one.) Tom’sversion does have a bit more detail that agreeswith Jim’s recollections. In particular, it seemsthat the first two rubber fowl purchased were onechicken and one turkey. Tom also comments, “Irecall the salesman at the joke shop pointing out,with particular pride, the authenticity of the chick-en: it had a “hickey” on the side of its neck, whichis present in properly killed chickens.” RobertMcCartney adds that the two versions are compat-ible, since Tom’s concentrates on the actions lead-ing up to the purchase of the chicken and Jim’s onwhat happened thereafter. Also, it seems plausi-ble to Robert that it was Jim who was awarded thefirst rubber chicken. He thinks Mark Post (PhD’85), also a student then, may have received theturkey. Robert’s (almost) final line was “PS—isthis how religions start?”Robert’s actual final comment came in a subse-quent email in response to my thank-you note,
t! 18
Michael Benjamin 2002 Post-doc at MIT’s Dept. of Ocean Engineering and AI Lab. He is developing AI techniques for autonomous navigation of unmanned marine vehicles in joint work with the Navy in Newport, RI
Stina Bridgeman 2002 Assistant Professor, CS, Colgate University
Steve Dollins 2002 Steve’s setting up an investment advisory business to manage stock portfolios. He hopes to be up and running by early next year
Anthony Fang 2003 Assistant Professor, CS, National University ofSingapore
Sam Heath 2003 Working for McKinsey and Co. in Toronto, doing management and strategy consulting for a variety of large companies, including,biotech, telecom, and (since this is Canada) natural resources
Kee Eung Kim 2001 Working for Samsung SDS Co., Ltd. in Seoul, Korea
James Kurien 2003 Researcher at Palo Alto Research Center, developing AI techniques for control of complex machines
Luis Ortiz 2002 Post-doc at the University of Pennsylvania
Gopal Pandurangan 2002 Assistant Professor, CS, Purdue University
Leonid Peshkin 2002 Post-doc research fellow at CSAIL, MIT’s Computer Science and AI Lab, working under LeslieKaelbling’s guidance once again. His work involves some student supervision and research into com-puter vision and text understanding, stressing the learning aspect of intelligence as opposed toengineering intelligent behaviors
Bill Smart 2002 Assistant Professor, CS, Washington University, St. Louis
Srikanta Tirthapura 2003 Assistant Professor, Electrical and ComputerEngineering, Iowa State University
which read: “Thanks! I now know more aboutrubber chickens than I ever thought possible. Ithink the last five brain cells I will ever createhave been dedicated to the topic.” I found hiscomment touching, particularly since it comesfrom the ever-cynical Robert McCartney:
Subject: RE: Rubber ChickensFrom: Robert McCartney<[email protected]>To: Eugene Charniak <[email protected]>Cc: Robert McCartney <[email protected]>eugene--one more cell’s worth: the most importantfact about the rubber chicken tradition isthat these are given to graduates by theirfellow students--not the department, not
condui
the advisor, not the university, not even theformer mayor of providence. this was true from the start (as was the casewith jim), and is perhaps the most appeal-ing aspect of the tradition.cheers, r.
Moving on to other topics, just after the last con-duit! came out, our chair, Eli Upfal sent out anemail with a very odd spelling error. The only per-son to notice it, however, was Shriram Krishna-murthi, and as the email was otherwise run of themill, I did not save Eli’s original. However, I surenoticed Shriram’s response:
Eli, while no-one would begrudge youyour spell-checked emails, you shouldconsider what the spell-checker is doingto your signature. Shriram
t! 19
Department of Computer ScienBrown University
Box 1910, Providence, RI 02912,
conduit!
Printed on recyled paper
cThe Co
DeB
WW
Eli Upfal wrote:Please vote for faculty committees. To-morrow is the deadline - some of our fac-ulty are on the ballots. You don’t need tovote for all categories if you don’t knowthe people.=============================
Eli Cupful, |Professor and Chair |Computer Science | (401) 863-7602Brown University | Fax: (401)863-7657Box 1910 | E-mail: [email protected], RI 02912 | http://www.cs.brown.edu
I asked Shriram why he spent time readingthe fluff at the bottom of emails. He said
ce
USA
conduit! 2
onduit!A publication of
mputer Science DepartmentBrown Uiversity
Inquiries to: conduit!partment of Computer Scienceox 1910, Brown University
Providence, RI 02912Fax:.401-863-7657
Phone:.401-863-7610email:[email protected]:.http://www.cs.brown.edu/
publications/conduit/
the conduit! teamanniversary. l
Eugene
that he doesn’t but that the fact that the “|”sin the signature no longer lined up caughthis eye.At any rate, this reminded me of a talk thatformer faculty member Dan Lopresti gavehere many years ago with the catchy title,“The computerless office.” His idea was thatpaper is the world’s most convenient storagemedium, and we should make it more com-patible with computers. In particular, sincethese days most, if not all, paper documentsare produced by computers, there is no reasonwhy they could not include some extra codedinformation that would help an optical char-acter recognition system reacquire the printedinformation. As one experiment Dan forcedhis OCR system always to produce actual En-glish words. He scanned in the first page ofMoby Dick and got the immortal first line:“Call me fishmeal.”
0
NON-PROFITU.S. Postage
PAIDProvidence, RI
Permit #202
Address changes welcomed
salutes the department’s 25th to r: Jeff Coady, Suzi Howe, Charniak, Trina Avery