GRAPH DRAWING and INFORMATION VISUALIZATION

conduconduit t! Volume 8, Number 1 Department of Computer Science Spring, 1999 Brown University

Traditionally, information visualization GRAPH DRAWING and has focused on the display of quantitative information (e.g., bar charts, pie INFORMATION VISUALIZATION charts, function plots) and geographic information (e.g., road and subway Information visualization is maps), where a natural mapping exists an emerging discipline that between the data and their geometric addresses the problem of location in the diagram. A great introduc- communicating the struc- tion to these types of visualizations is ture of information space given in the books by Edward Tufte (The through diagrams. Quot- Visual Display of Quantitative Informa- ing from an article by David tion and Envisioning Information). Harel (Comm. ACM 31(5) 1988): From quantitative to relational “The intricate nature of More recently, researchers have started a variety of computer- addressing the problem of visualizing related systems and sit- relational information, where net- uations can, and in our works (also known as graphs) model opinion should, be rep- collections of objects and connections resented via visual for- between those objects. Examples include: malisms; visual because • Web: site maps, browsing history dia- they are to be gener- grams, presentation and refinement ated, comprehended, of query results, product catalogs. Roberto Tamassia and communicated by humans; and attending a workshop formal, because they are to be manip- • Software engineering: UML class and state-transition diagrams, sub- on graph drawing in ulated, maintained, and analyzed by Barbados computers.” “...in the next decade the use The benefits of information visualization include of information visualization analysis through visual exploration, discovery of techniques will be essential to patterns and correlations, the success of portals and other and abstraction and sum- marization. It is antici- large information repositories pated that in the next decade the use of informa- on the Internet” tion visualization techniques will be routine-call graphs, data-flow dia- essential to the success of portals and grams. other large information repositories on • Database systems: entity-relation- the Internet. ship diagrams.

Brown University, Box 1910, Providence, RI 02912, USA • Real-time systems: Petri nets and To complicate matters, some basic graph state-transition diagrams. drawing problems for which theoretically • Networking: LAN diagrams. fast algorithms are known turn out to be • Enterprise and project manage- unwieldy to implement. Take, for in- ment: business process diagrams, stance, the problem of testing whether a organization charts, scheduling graph is planar, i.e., whether it can be charts. drawn without crossings. While mathe- • Engineering: circuit schematics. matical characterizations of planar • Artiﬁcial intelligence: knowledge graphs have been known since the 18th representation diagrams, belief and century, it was only in 1974 that John inﬂuence networks. Hopcroft and Robert Tarjan published the first linear-time algorithm to test The fundamental problem in the visual- whether a graph is planar (J. ACM ization of relational information is the 21(4)). This algorithm was a major theo- automatic layout of networks, which is retical accomplishment and greatly con- the subject of the research area known as tributed to their earning the prestigious graph drawing. Several pioneering Turing Award. commercial applications have begun to appear. For example, the AltaVista Coming up with a correct implementa- search engine by DEC/COMPAQ can tion of the algorithm was, however, a dif- visualize the results of a query with an ferent matter. The intrinsic conceptual automatically generated drawing of a difficulty of the approach combined with graph whose vertices are relevant key- data-structuring tricks and special cases words, and supports the refinement of defeated the efforts of many program- the query by a direct manipulation of the mers for more than 20 years until finally drawing. The Hyperbolic Tree (TM) tech- in 1996 a research team led by Kurt nology for drawing trees by Inxight Soft- Mehlhorn completed the development of ware (part of the Xerox New Enterprise a reliable implementation of the business initiative) is used in the Web Hopcroft-Tarjan planarity testing algorithm that has been successfully tested on tens of thousands of graphs. Graph drawing research at Brown Since my first paper on automatic layout of entity-relationship diagrams was published in 1983, graph drawing has been one of my main research interests. Three of my six doctoral students to date have done their research on graph drawing:

AltaVista Web site of the Wall Street Journal (go to search display the “Money Tree”) and is incorporated in a Web management tool by Microsoft. There is significant potential for enhanc- ing electronic commerce Web sites with graph-drawing technology. Bob Cohen Ashim Garg Robert Cohen (Ph.D. 1992, now at Algo- While the task of automatically produc- magic Technologies, Inc.), Ashim Garg ing a readable layout for a graph may (Ph.D. 1995, now at SUNY Buffalo), and appear simple to a nonexpert, it is actu- Stina Bridgeman (current; a piece on her ally computationally very hard. The cost research appears in this issue). of incorporating effective automatic network-layout capabilities into software My recent work on graph drawing has systems is often grossly underestimated. focused on:

conduit! 2 • interactive layout techniques includes the development and commer- • Web-based graph-drawing systems cialization of a package of Java software • software design patterns for graph components for graph layout in collabora- drawing tion with Algomagic Technologies, a recent startup founded by Robert Cohen, Graph drawing on the Web Michael Goodrich from Johns Hopkins The Graph Drawing Server is a Web- University and myself. based system that provides graph drawing services. It can be accessed in vari- A book on graph drawing ous ways: I have recently published a book on the • through an interactive graph editor subject in collaboration with three other implemented as a Java applet graph-drawing gurus: Giuseppe Di Bat- tista from the University of Rome, Italy, • through an HTML form that allows Peter Eades from the University of New- the definition of the input graph in a castle, Australia, and Ioannis Tollis from variety of formats the University of Texas at Dallas. The rig- • through a Java package that pro- orous treatment of the subject allows this vides an API for involving the server book to be used as a text for graduate It supports various layout styles and drawing algorithms, including orthogonal layouts computed by the GIOTTO algorithm and hierarchical layouts

Graph-drawing gurus. l to r: Ioannis Tollis, Roberto, Giuseppe Di Battista and Peter Eades courses and as a reference for researchers, while the large number of examples and ﬁgures makes it also suitable for soft- Graph drawing server ware practitioners. The latest issue of Dr. Dobb’s Journal computed by an algorithm that distrib- (June 1999, page 134) mentions the book utes vertices on horizontal layers, and as follows: has been successfully used by many researchers worldwide to experiment “The final book this month is with graph-layout techniques. GRAPH DRAWING: Algorithms for the Visualization of Graphs,by Graph drawing in Java Giuseppe Di Batista, Peter Eades, Roberto Tamassia, Ioannis G. Tollis. I am developing a library of reusable The title is an accurate summary of software components for graph drawing the book’s contents, but doesn’t do in Java. This work is based on algorithm justice to its breadth. Section 5.1, for engineering techniques that include the example, is devoted to angles in use of novel algorithmic patterns. The orthogonal drawings, while chapter library can be used to incorporate auto- 7 covers incremental construction matic layout capability in various inter- technniques. The style is academic— faces that make use of diagrams. I plan there are a lot of references, and a to show its applications especially to Web lot of proofs and lemmas—but the browsers and programming environ- book will be a rich mine of ideas for ments. anyone who is trying to persuade a In addition to algorithmic and systems computer to turn data into dots, research, my future graph-drawing work boxes, lines and arrows.”

conduit! 3 “HORIZON” This imposing iron and azure glass monument by artist Costas Varotsos was commissioned by General and Mrs. Kanellakis in memory of their son Paris and his family. With majes- tic Mt. Parnassos as a back- drop and the town of Liya below, the sculpture is located on family-owned land—a favorite summer haunt where they’d gather each year to take family snapshots. The inscription reads: DEDICATED TO OUR CHILDREN PARIS - MATE - ALEXANDRA - STEPHANOS THEIR PARENTS ELEFTHERIOS AND ROULA KANELLAKIS 20-12-95 Paris’s legacy will endure via fellowships and awards established to honor his memory—the ACM’s Kanellakis Award, Brown’s Kanellakis Graduate Fellowship and MIT’s Kanellakis Graduate Fellowship, to name a few. His parents have donated the land around the monument, aptly, to the SOS Children’s Village International Parc. Said his mother, “It was their favorite summer place on earth and now their spirits shall dwell there—near to us—while those we cherish are young and together in eternity.”

using the technologies effectively and/or THE UNIQUENESS OF CS92 on a large scale. He writes that “claims predicting extraordinary changes in teacher practice and student learning, The failure of contemporary mixed with promotional tactics, domi- technologies significantly to nated the literature in the initial wave of transform, much less im- enthusiasm for each new technology. prove, elementary and sec- Seldom were these innovations initiated ondary education in the 20th by teachers” (p. 4). century is by now well- known. Even if radio, televi- Cuban also described a cycle that has sion, film, and now comput- characterized the fate of classroom tech- ers have all contributed a nology since 1920: exhilaration, followed great deal to changing what by scientific credibility, then disappoint- we teach and learn, they ment, and finally teacher-bashing. First thus far have had remark- published in 1986, Cuban’s book had lit- ably little effect on how we tle to say, and no interesting data, about teach and learn it in school. the use of computers, but anyone follow- Roger Blumberg ing even newspaper articles about com- In his insightful book, Teachers and puters in schools over the past few Machines (Teachers College Press, months will realize that all the stages of 1986), the historian Larry Cuban pre- Cuban’s cycle can already be identified sents data suggesting that, among other in the short history of computers as a things, the failure of classroom technolo- species of classroom technology. gies in American schools in this century In its preference for top-down design and has had much to do with a top-down promotion of educational technology, a method of implementation and promo- good deal of university research has only tion which has kept school teachers from contributed to the unfortunate story of

conduit! 4 computers in school. The typical univer- of the Seminar have been—a phenome- sity-school collaboration model in com- non due in no small part to the excellent puter science has been that a group of technical preparation the students university researchers develops a tool, or receive in the CS Department—it is the set of tools and curricula, based on their process that, in my view, makes the Sem- own expert intuitions of good design, inar a valuable, liberal arts (i.e. non- instructional need, and educational val- technical) undergraduate experience. ue, and then attempts to recommend the The Educational Software Seminar has tool, and the vision implicit in its design, been in existence for more than a decade, to working teachers. This model has met and it owes that existence to Andy van with remarkably little success, consider- Dam. Orchestrating and teaching the ing the tremendous expertise (technical Seminar with David Niguidula in the and pedagogical) and resources (public early and middle 1990s, when students and private) that have been devoted to it. built software in Hypercard for teachers in schools with Macintosh computers In light of the history of classroom tech- (small in size and number), Andy devel- nology, the difference in the approach to oped a curriculum for the course and his the study and use of computers at school students produced a set of programs that embodied in The Educational Software remain interesting and useful in the study of educational software. Indeed, “The Seminar begins not many of those programs are still popular with a product, deduced with teachers and students—a great tribute both to Andy’s and David’s idea of from first principles, but with the course and to how much their students were able to do even with very proposals gathered from primitive authoring tools. working teachers” Last year I began teaching the Seminar, and decided first to expand the pool of Seminar at Brown (CS92/ED89) is clear. possible projects to include proposals not The Seminar begins not with a product, only from teachers in K-12, but from deduced from first principles, but with Brown University faculty and educators proposals gathered from working teach- in other community institutions as well. ers. These teachers specify instructional The range of possible platforms was sim- objectives and technical constraints, as ilarly expanded to include not only the well as their speculations and aspirations PC but the Web as well. With the kind concerning how the computer might and indispensable cooperation of the improve student learning and their own Multimedia Lab at Brown, CS92 stu- teaching. dents were able to choose from a variety In the spring semester, Brown under- of multimedia authoring tools, in addi- graduates enrolled in the Seminar read tion to the programming languages these proposals—which are solicited, available and taught in CS, and thus gathered and negotiated in the fall—and could design for a greater number of choose those in which they are most platforms, instructional goals, and tech- interested. Then, working in teams of nical constraints than in previous years. three or four, students in CS92 design, Finally, the readings for the course were create, test and implement the programs revised to incorporate recent studies like proposed. In the process of building the Cuban’s and the 1997 analysis of the software, the teams work closely with Apple Classrooms of Tomorrow Project their sponsoring teacher and her/his stu- (Judith Sandholtz et al., Teaching with dents, and the production work (which Technology [Teachers College Press, includes the creation and presentation of 1997]), and to better reflect the interdis- storyboards as well as prototypes) is done ciplinary nature of the Seminar. in parallel with seminar discussions of For all the changes, however, Andy’s and literature drawn from computer science David’s idea of the Seminar remains, and and the history and philosophy of educa- this is what makes CS92 such a success- tion, as well as the cognitive and social ful course and worthwhile experience for sciences. As remarkable as the products students as well as our clients. Last year

conduit! 5 our students worked with elementary, students, some of them reading philoso- secondary, university and community phy, cognitive science, and/or social sci- teachers, creating programs for settings ence for the first time at Brown in CS92, as diverse as the Blessed Sacrament are able to arrive at important, novel School (with Macintosh computers insights about the connections among equipped with only 4MB of RAM) and “...computer science students...are the Brown School of Medicine; covering able to arrive at important, novel subjects as different insights about the connections as 4th grade social studies and under- among questions concerning tech- graduate visual art; nology, pedagogy and cognition” using tools as various as Hyperstudio and Java; and serv- questions concerning technology, peda- ing audiences as diverse as 7th grade gogy and cognition. science students at the Wheeler School In his important book, The Language of and visitors to the City Streets exhibit at Education (Charles Thomas, 1960), the the Providence Children’s Museum. All philosopher Israel Scheffler wrote that of the programs were completed and con- “educational research must not be con- ceived as a single science, but rather as the common focus of many sciences with bearings on educational practice” (p. 73). While I won’t raise local blood-pressures by making a similar claim about computer science research, the experiences of the students in CS92 clearly demon- strate how success in the creation of educational software necessarily involves ideas from the computer, cognitive and human sciences, as well as creative insights not clearly derivable from the first principles of any (known) science. This sort of non-algorithmic complexity makes the creation of genuinely effective educational software extremely difficult, and indeed the computer industry has Visitors to the City tinue to be used, not only by their spon- largely retreated from education to Streets exhibit at soring teachers but by teachers and entertainment and from the classroom to the Providence students who have downloaded them the home; but, of course, it is precisely Children’s Museum from the course Web site as well. this sort of complexity that makes educa- This year, we are again working with the tional software, and educational technol- Vartan Gregorian Elementary School ogy in general, such a challenging but and both undergraduate and post-gradu- potentially rewarding area of research. ate teachers at Brown, and we are The Educational Software Seminar pro- excited to be working with the elemen- vides not only a unique, productive tary school classroom at Hasbro Chil- model of university-school and univer- dren’s Hospital for the first time. The sity-community collaboration, but an students in the Seminar are working important method of inquiry into the with teachers and their students, docu- relation between technology and educa- menting their activities on project pages, tion as well. CS92 projects and pro- thinking carefully about pedagogy, learn- grams, past and present, along with ing and the design of electronic media, team project pages and our syllabus and and are again doing remarkable work. bibliography, can all be found at the This year I’ve been especially struck by course Web site: http://www.cs.brown.edu/ the extent to which computer science courses/cs092/.

conduit! 6 Symposium speakers from l to r: Maurice Herlihy, Clare Rabinow, Lotus; Bob Morgan and Bill McKeeman, Compaq; Nikos Aneuris, AT&T Research; Ken Arnold, Sun; Bill Kayser, WorldStreet

registering itself with a Jini Lookup Ser- vice. A potential client can locate a THE 22nd IPP SYMPOSIUM server by matching a Java type, possibly augmented with additional attributes. The 22nd Industrial Partners Program Code moves from the service to the technical symposium, held on November Lookup Service to the client. Code need- 12, 1998, was on “Realizing the Poten- ed to use the service is dynamically tial of Java,” a topic selected for its loaded on demand from the client. Jini relevance to the current interests of makes no distinction between hardware many Partners. The speakers included and software. Services can be local, representatives from Partners Sun remote, or a combination. It accommo- Microsystems, Lotus/IBM, and Compaq, dates legacy services and languages. Jini as well as AT&T Research and World- is not an operating system, but eventu- Street. The topics covered fell into two ally could replace the operating system broad categories: technology to support with a ubiquitous and invisible network. Java (Sun’s Jini and Compaq’s ahead-of- time compiler) and experience using Clare Rabinow (Lotus), a Brown al- Host: Maurice Herlihy Java for critical enterprise-wide applica- umna (class of ’72), described her group’s tions (Jini, MARVEL, and eSuite). experience implementing eSuite, a business productivity application that en- Ken Arnold (Sun Microsystems) dis- compasses components for data presen- cussed the newly-released Jini system. tation and editing and data access, as Jini is described as “network plug-and- well as traditional business productivity work”: it is intended to ease the distinc- applications for the network computer. tion between hardware and software by The talk highlighted both the advan- allowing spontaneous networking (con- tages and the pitfalls of using Java for necting components on the ﬂy using a an enterprise-level application. The prin- simple common interface). Java itself cipal advantages include the experience provides a homogeneous network in the that it really is feasible to use Java to form of safe, portable object code and a develop software for dissimilar plat- single type system for the entire net- forms; it is relatively easy for customers work. Jini extends this model by to mix components from different ven- permitting a service provider to adver- dors; and businesses can build applica- tise its service across the network by tions that beneﬁt from platform inde-

conduit! 7 pendence and low-cost deployment with- protocol adapters to interact with compo- out ever learning to program in Java. nents using both standard and pro- Among the pitfalls, Rabinow observed prietary management protocols. that every Java Virtual Machine has its Bill Kayser (WorldStreet) described his own bugs and quirks, and every runtime company’s experience building a produc- environment its own schemes for secu- tivity system providing real-time market rity and for locating language resources, information for securities trading. The and that Browser/Java Virtual Machine tool described integrates real-time mar- deployment lags behind availability. The ket status, a chat-style applet for group principal performance bottleneck turned communication, and integrated access to out not to be execution speed, which is database and enterprise information. more than adequate for interactive appli- Java provided substantial advantages in cations, but delays in launching ap- terms of programmer productivity, partic- plications and downloading classes. The ularly due to exception handling, garbage absence of pointers and explicit memory collection, and the rich collection of class deallocation do eliminate many tradi- libraries. Experience with performance tional sources of memory leaks, but leaks was mixed: WorldStreet encountered a still occur in the form of references “bur- wide variability in Java virtual machine ied” within tables and Abstract Window implementations, and garbage collection Kit peer resources. introduced some scalability problems. The principal hazards encountered in- Nikos Anerousis (AT&T Research) cluded spaghetti code introduced by care- described MARVEL, an object-oriented less use of exception handling, thread- Java-based toolkit for deploying scalable related deadlock and performance prob- network management services. The arch- lems, and unexpected behavior of static itecture has two core components: an initializers. Development suffered from information model that allows network the lack of several kinds of tools: debug- management information to be aggre- ging, platform-speciﬁc development tools, gated in spatial, temporal, and functional and conﬁguration management. forms, and a presentation model that interacts with Web browser clients to Robert Morgan and Bill McKeeman launch Java applets to perform data dis- (Compaq) reported on Compaq’s opti- play. Marvel is based on a client-server mizing “ahead-of-time” Java compiler. structure. Clients are simple Web brows- They observed that generating good code ers and servers for advanced processors is expensive, “I think that I shall never see are extensible ag- since one must deal with software pipe- Such tortes as those for IPP. Naught compares to cakes like that, oh ents that employ lining and instruction scheduling, reg- How I long for chocolate gateau!”

...eat your heart out, Melvin Belli

Besides distinguished speakers, IPP symposia are noted for an abundance of lavish desserts

conduit! 8 ister allocation, loop transformations, structure that starts with Java byte and a host of similar optimizations criti- codes and successively either computes cal for exploiting the power of modern information for later phases or moves the processor architectures. These kinds of program closer to machine instructions. optimizations cannot be done effectively The ﬁnal result is a “native” machine- by a just-in-time (JIT) compiler because code representation of the program, they take too much time. An ‘‘ahead-of- encompassing a level of optimization time’’ compiler requires advanced analy- beyond what a just-in-time compiler sis to generate good code. Challenges could provide. include computing concrete types, opti- This IPP symposium gave an rare snap- mizing dynamic method calls, placing shot of how well Java is realizing (and data objects on the stack whenever possi- sometimes not completely realizing) its ble, eliminating null reference and array potential to transform the computing bounds checks, and analyzing aliasing world in essential ways. among references, as well as classical optimizations. They described a compiler

course, tradeoffs inherent in this— INTERACTIVE GRAPH DRAWING requiring a straight-line drawing may increase the area requirements, minimiz- and SIMILARITY METRICS ing the area may increase the number of edge bends, and so forth.

Interactive Graph Drawing A great deal of work has been put into developing graph drawing algorithms and one can find an algorithm that does a good job of layout for a variety of applications. Many of these algorithms, however, were developed using a batch model in which the graph is redrawn from scratch each time. Such algorithms are not well suited for interactive applications, where the typical scenario is that the user requests an initial drawing of the graph, then makes some small changes and requests a new drawing, and then makes another small change or Stina Bridgeman Being able to draw nice pictures of two and requests a third drawing, and graph-structured information is a useful then makes yet another change... These changes may be made because the user is thing—data structure visualization, vis- explicitly editing the graph or because ual programming, database design, Web the graph represents some structure that navigation, and a variety of other appli- is being updated, such as a map showing cations can benefit from clear and the user’s navigation through a collec- understandable drawings of graphs. A tion of Web pages or topics returned by a graph-drawing algorithm typically at- search engine. tempts to optimize certain aesthetic cri- In this interactive model, it is assumed teria while working within a given that the user has gained some knowl- drawing paradigm. So, for example, an edge—built up a “mental map”—of the algorithm may attempt to minimize the previous drawing, and so it is desirable number of bends in an orthogonal draw- to preserve the mental map by having ing, where vertices are placed at grid small changes in the graph structure points and edges are chains of horizontal translate into small changes in the draw- and vertical segments. There are, of ing, even at the expense of the other

conduit! 9 An orthogonal drawing, with New drawing produced by some user changes InteractiveGiotto ings of graphs? Interactive drawing algorithms frequently seek to preserve the mental map by attempting to minimize the change between drawings, typically by allowing only very limited modifications (if any) to the position of vertices and edge bends in the existing drawing. Some layout adjustment algorithms (where the idea is to rearrange a drawing in order to improve some aesthetic criteria) use a notion of proximity to preserve the mental map, by requiring that a point’s position in the new New drawing produced by Giotto drawing be closer to its own old position aesthetic criteria. Redrawing the graph than any other point. Orthogonal order- from scratch at each step often causes ing has also been proposed as a factor in large changes in the drawing, destroying similarity—if the relative north/south/ the user’s mental map and forcing her to east/west relationship between pairs of spend considerable time refamiliarizing vertices is preserved, the new drawing herself with the layout—imagine how dis- will tend to look more like the original orienting it would be if the map of the than if these relationships are changed. Web pages you’ve explored were com- The goal of my current research is to pletely redrawn each time you visited a define and evaluate similarity metrics for new page. graph drawings, and in so doing, gain Of course, eventually the “at the expense some insight into what features of the of the other aesthetic criteria” bit adds up drawing are most important for visual and it may be desirable to redraw the similarity. Such insight will be useful in graph from scratch periodically, but a designing new interactive algorithms good deal of effort on the user’s part can that better optimize the desired aes- still be saved when this is done only peri- thetic criteria while preserving the odically and not after every change. mental map. Developing similarity metrics will also provide a basis of comparison between different drawing algo- Similarity rithms—currently the comparison is still The goal of preserving the user’s mental done in terms of “traditional” optimiza- map brings up the idea of similarity—if tion criteria like the area and the the new drawing looks similar to the old, number of edge bends, and discussion of the user’s familiarity with the old draw- mental map preservation is limited to ing will transfer to the new drawing with stating that the algorithm does a good minimal effort. The question is, how does job because it doesn’t change the draw- one measure similarity between drawing much.

conduit! 10 My focus has been on measuring similar- problem comes up because InteractiveG- ity between two drawings of the same iotto throws away coordinate information graph; determining how similar two dif- when computing the new drawing, and ferent graphs are is another problem thus may produce a drawing that is entirely, and one that is less immedi- rotated by a multiple of 90 degrees with ately applicable to interactive graph respect to the original. A metric that drawing. The initial step has been to chooses the correct rotation can be used define metrics based on mental map to fix InteractiveGiotto’s output, which is models that have been proposed in the important because a large rotation is a literature—vertex position, proximity, very easy way to destroy the user’s men- orthogonal ordering and edge shape. tal map. (For orthogonal drawings, edge shape is To evaluate the metrics it is necessary to the sequence of left and right turns made obtain multiple drawings of the same along each edge.) graph. For orthogonal drawings, Interac- A successful similarity metric should tiveGiotto is well suited for this purpose pass three tests. It should qualitatively because its constraints can be relaxed on reflect human judgment so that ordering a vertex-by-vertex or edge-by-edge basis, a set of new drawings with respect to the making it possible to obtain multiple original based on the metric’s value drawings of the same graph with vary- yields the same ordering a human would ing degrees of similarity to the original produce when asked to arrange the same drawing. (This is in contrast to most drawings. It should also quantitatively interactive drawing algorithms, which produce a single output drawing for a given input.) Having multiple drawings “A metric that chooses the cor- of the same graph is important because it rect rotation can be used to ﬁx is very difficult to judge whether one pair of drawings is more or less similar than InteractiveGiotto’s output, which another when the two pairs are draw- is important because a large ings of different graphs. So far I have focused primarily on evalu- rotation is a very easy way to ating the metrics based on their destroy the user’s mental map” qualitative behavior and their ability to choose the correct rotation, because hav- reflect human judgment so that the ing a human assign meaningful numeric value of metric is proportional to the per- similarity values to drawings is very dif- ceived difference in similarity. And ficult. (One future project is to devise a finally, it should choose the correct rota- way to get such values in as objective a tion; namely, given the original drawing manner as possible.) There have been and a set of new drawings identical to some winners and some losers—the each other except for a rotation factor, orthogonal ordering metrics tended to do best in both respects, with the proximity the metric’s value should be the lowest metrics faring the worst—but nothing so for the rotation that a human would say far has been an obvious “perfect metric.” best matches the original drawing. This Most of the metrics are suitable for last condition is the easiest to satisfy, choosing the correct rotation, at least and was motivated by the “rotation prob- when the possible angles of rotation are lem” of InteractiveGiotto. widely separated (e.g., multiples of 90 InteractiveGiotto, largely developed by degrees), and most of the metrics do an Jody Fanto ’97, is essentially a front-end OK, but not stellar, job of ordering the to the successful (batch-model) orthogo- drawings. nal drawing algorithm, Giotto. It seeks This may be because these metrics all to preserve the look of the drawing by consider the big picture, taking the view applying constraints fixing the angle that all parts of the drawing are created between adjacent edges around vertices equal and so a change has the same and the number and direction (left or effect wherever it is applied. This is not right) of bends along edges. The rotation necessarily the case, though—while

conduit! 11 attempting to order the drawings for separate from other parts of the draw- evaluating the metrics, I quickly realized ing. My current direction is to further that landmarks are important. One tends investigate how to identify landmarks, to focus quickly on sections of the draw- how to measure their importance and ing that are distinctive, and if these are how to measure how they change, in the present (and relatively unchanged) in the hopes of improving the similarity met- new drawing, it greatly increases the per- rics. ceived similarity of the drawings. Also, And if anyone has a desire to spend time landmarks vary in importance—a very staring at lots of drawings of graphs, distinctive structure in the graph is more come see me. :) noticeable, as is one that is relatively

exactly a household name, but at the Uni- versity of Washington, where he is chair LETTERS & ALUMNI EMAIL of Computer Science and Engineering, and among state education leaders, he is E. GORDON GEE, President something of a hero. Lazowska has assist- Dear Suzi—A note to tell you that I just ed Seattle Public Schools in a variety of read the latest issue of conduit!.Itisa ways, from Internet connections to tech- ﬁrst-rate publication and certainly rep- nology training for teachers. He is a resents the Department, as well as the leader in the K-20 network that will University, very well. Would you please someday link all the state’s schools and pass my delight in that product to others colleges. Lazowska was a 1998 recipient who contributed to this? Best personal of a UW Public Service Award. He is rec- wishes. ognized for his efforts to bolster the state’s economic future through technol- JASON LANGO AND NATE STAHL, ogy-smart policies.” ScBs ’98 Hello friends and former colleagues! It EVAN MAIR, BA ’93 would take a while to tell you all individ- As the popular saying goes, you can take ually, so I thought that I would be lazy the Brown CS grad out of the Sunlab, but and spam you all. Nate and I have de- you can’t core-dump the desire to code. cided to leave Silicon Graphics and pur- During my years in medical school, I sue career opportunities at a much developed commercial medical software smaller company named Network Appli- for the 3Com Palm series as well as help- ance (www.netapp.com) working on ing the Boston University Medical Cam- systems software for the NetCache box. pus develop its Internet presence. I am We’re unable to check our email at SGI, now entering my residency in Diagnostic but the mail addresses [email protected] Radiology at Boston Medical Center, and [email protected] will continue to where I hope to continue my research on reach us at our apartment. image compression and archiving systems for radiology services. I can be ED LAZOWSKA, AB ’72 reached at [email protected]. The following was sent to Andy van Dam by David Salesin, ScB ’83, under the sub- LAURENT MICHEL, PhD ’99 ject line “Remarkable people:” Hi! I am now working for Ilog S.A., a “Every Jan 1, the Seattle Times recog- French company that specializes in the nizes ‘remarkable people...who added production of software libraries for combi- measurably to the quality of life in the natorial optimization problems. About a Puget Sound region.’ Among the half- year ago they acquired CPLEX (the dozen individuals who are recognized this leader in linear programming). Basically, year is our own Ed Lazowska. Read it Ilog has a suite of products ranging fro online at http://www.seattletimes.com/ libraries for LP/IP/MIP to constraint pro- news/editorial/html98/bested_010199.- gramming libraries (general-purpose CP) html, where they say “Ed Lazowska isn’t and even scheduling. They also have a

conduit! 12 whole different sector of activity in com- to the Bay Area. I am leaving the computer graphics (high-performance 2D/3D fortable and laid-back environment of rendering) and telecom. Bell Labs for the thriving madness of Sili- I am working in the Optimization R&D con Valley. I am joining a small company team in the modeling language group. I called Epiphany. We make ERM (enter- am working from Belgium while Ilog prise relationship management) software. headquarters are in Paris. You can ﬁnd I would love to hear from friends. I can be out more about them at http://www- reached at [email protected]. .ilog.com. My work email address is [email protected], and I’m more likely to answer quickly on this one! MICHAEL RUBIN, ScB ’99 Hello conduit! It’s funny how only after VISWANATH RAMACHANDRAN, I leave Brown do we get a chance to talk. PhD ’98 Currently I am in California working at After my Ph.D., I joined Netscape (now a Network Appliance and having a great unit of AOL), working on the Netscape time. Web server as an engineer. I’ve been After all the time and all the work, I focusing mostly on programming lan- think I am enjoying the fruits of my guage work, developing language run- labors—a job where I sit in one room all times for Java and JavaScript. I am day pushing buttons. Seriously, I am hav- really excited about being in the forefront ing a great time out here. It’s amazing of the NetEconomy and electronic com- how much easier it is to work without merce. On the personal front, I got sleep deprivation. It’s hard to write one married to Shanthi in April 1998. Stay in of these without sounding like all the touch folks, my email is vishy@netscape hundreds of other recent grads who have .com. nothing monumental to say. Work is better than I hoped. Love is better than I SRIDHAR RAMASWAMY, PhD ’98 hoped. Life is good. Thanks to all those Hi everyone! My wife Seema (Tufts DMD who helped make this happen, mhr. ’96) and I have moved from New Jersey [email protected].

Unfortunately, about two months before [email protected] the conference Eugene had a skiing accident, and his physical therapist assured Eugene Charniak. This last year him that he would be crazy to try skiing Eugene Charniak gave two invited talks given the state of his left knee. So, while at conferences on machine learning. He his talk was very well received, the con- was pleased about these invitations ference was less than a total success for because, while his research has always him. been in natural-language processing, it At the beginning of this last February was only a few years ago that he began to Eugene (as well as Tom Dean) went to a apply machine-learning techniques to the conference center in Conway, CT, a noted problem. Thus he was happy to get this ski center, for a retreat for students and indication of “acceptance” in his new professors connected with some new research community. Of these invitations, interdepartmental grants (Computer Sci- the more prestigious was from the NIPS ence, Neuroscience, Cognitive Science, (Neural Information Processing Systems) and Applied Math). Again, afternoons conference. However, the invitation that were free for skiing. With a year of knee initially had him more excited was the exercises under his britches, Eugene ven- Snowbird conference on machine learn- tured onto the slopes and is pleased to ing, a conference famous because it is report that his knee held up just ﬁne. held in ski season at a ski resort, and afternoons are left open for skiing. Eugene always felt that clearly these ▼▼▼ folks knew how to throw a conference!

conduit! 13 Tom Dean. Tom, program committee John Savage. In April John served on chair for IJCAI-99, attended a meeting of a committee that wrote a report to the the IJCAI Inc. board of trustees in Paris NSF entitled “Challenges for Theory of during the World Cup Games. Computing.” This report offers NSF advice on funding research in theoretical computer science. Its URL is http:// ▼▼▼ www.cse.buffalo.edu/~selman/report.

▼▼▼ John Hughes. Well...the graphics group sent six papers to SIGGRAPH— actually, something like a total of 12 items went out, from course and panel Roberto Tamassia. Roberto cochaired proposals to papers to who-knows-what. the program committee of the Workshop One cool thing was the paper-that-be- on Algorithms and Data Structures (Van- came-a-paper in the last three days couver) and served on the program com- before the deadline, with two undergrad- mittee of the Workshop on Algorithm uates as lead authors. As for me person- Engineering and Experimentation (Balti- ally, I’ve been working on the Andyfest more). He also gave a keynote lecture at (the May 27-28 gathering of Andy’s the Symposium on Algorithms and Com- friends, colleagues and ex-students world- putation (Taejon, South Korea). wide to celebrate his 60th birthday), my courses, the ﬁve-year review of the grad ▼▼▼ program and being graduate advisor—all the stuff professors do.

Eli Upfal. Eli’s financial computing ▼▼▼ research group visited Goldman Sachs in New York in December. He participated in an NSF review panel in January and Franco Preparata. Franco was ap- was appointed associate editor for the pointed to the international committee SIAM Journal on Discrete Mathematics. that assigns the Gödel Prize in theoretical computer science. He was also appointed to the Scientific Board of ISTI ▼▼▼ (Istituto per le Scienze e la Tecnologia dell’Informazione) in Pisa, Italy, a unit of the Italian National Research Council. In Andy van Dam. Andy has been the late fall of 1998, he successfully led awarded the 1999 IEEE James H. Mulli- the technical review of the project “Ro- gan, Jr. Educational Medal “for his field- bust and Applicable Geometric Comput- defining textbooks, the introduction of ing,” which, funded by ARO, is the main innovative educational technology, and support of our three-university Center for inspired undergraduate teaching.” Geometric Computing. This review cap- ped the initial three-year phase and has secured funding for the subsequent two- ▼▼▼ year option. At RECOMB99, the premier international conference in computational biology (held this year in Lyon, Peter Wegner. Peter is retiring in France), he presented the results of his June. This spring and summer he will be work with Eli Upfal, which are the sub- lecturing in Amsterdam, Lisbon and ject of a recent patent application. Tokyo. He was recently awarded the Aus- trian Medal of Honor in Science and Art. ▼▼▼ ▼▼▼

conduit! 14 The Dome, MIT and The Bridge of Sighs, St. John’s College Smith, who designed the graphics effects FROM CAMBRIDGE TO CAMBRIDGE in Star Wars at LucasFilm and Toy Story at Pixar and is now a researcher at Mi- crosoft. When I told him I had asked Bill During the week of April 12-16, 1999, I Gates about the goals of graphics, he attended the 35th anniversary of MIT’s cupped his ear and waited expectantly. Laboratory for Computer Science and the On hearing that Gates had mentioned re- 50th anniversary of the Cambridge (En- alism as a major goal, Alvy said there is gland) Computing Lab. The MIT event much debate in the graphics community featured Bill Gates’ $20-million gift and about the role of realism in graphics re- keynote talk about the future of comput- search. In fact, the Brown graphics ing, and the unveiling of MIT’s Oxygen group’s project on non-photorealistic ren- project for next-generation intelligent dering specifically explores nonrealistic personal assistants. The British event, a abstractions that focus on details of a vi- gathering of about 500 mainly British sual image considered relevant while ig- computer scientists, featured a keynote Peter Wegner noring details considered irrelevant. talk by Maurice Wilkes, who developed the first stored-program computer in the late 1940s and retains his resilience at “Abstraction is the age 85. Since I cannot do justice, in this enemy of realism” short article, either to the goals of Oxy- gen or to the British views of the past Data visualization strays from realism and future of computer science, I shall even further, since its visualizations relate some anecdotes. have no realistic image to which they In talking to Bill Gates, I asked about correspond. Abstraction is the enemy of Microsoft’s goals for computer graphics. realism. Bill seemed surprised at the question, as At dinner with Maurice Wilkes and his though he had not thought about it be- wife Nina, we revisited the early history fore. He spoke of Microsoft’s wonderful of computing. Maurice discussed his trip researchers in the graphics area, and to the US in the late 1940s to attend a said that one of the goals of graphics was course that enabled him to build the first realism. A few days later, at the confer- stored-program computer in 1949, one ence dinner in St. John’s College in the year ahead of his US competitors. We other Cambridge, I sat next to Alvy Ray talked about my days as a student in the

conduit! 15 first ever computing course during 1953- academic career. These events marked 54. I was later given a copy of the 1954 not only a celebration of the past, but final exam, which I would certainly fail also a passing of the baton from the old today because it included hardware guard to a new generation who must deal questions as well as advanced numeri- with new technologies of networking, col- cal analysis. laboration, and electronic commerce. Oxygen is an ambitious effort to develop On the final day of the conference I was an integrated approach to technologies of invited to lunch at the Microsoft Re- the 21st century, but foundations for in- search Lab by Roger Needham, who tegration have yet to be developed. headed the Cambridge Computing Lab before Robin Milner took over three The Microsoft connection at both Cam- years ago. It occupies a very pleasant bridge reunions reflects Microsoft’s emer- space in the center of Cambridge and gence as a ubiquitous presence in com- has some impressive researchers, includ- puting on both sides of the Atlantic. Mi- ing Luca Cardelli and Tony Hoare, as crosoft’s attention to the past could well well as a superb panorama overlooking prove useful in its bid to define the fu- one of the main thoroughfares. ture. If Microsoft succeeds, historians will identify the shift from old to new These two anniversary celebrations al- technology with a shift from an IBM- lowed me to meet many old friends who dominated to a Microsoft-dominated com- had traveled on parallel tracks and puter industry. were, like me, nearing the end of their

we’ll remember Markus for his generous FROM THE CHAIRMAN and enthusiastic spirit and his infectious sense of fun. This short note re- The next piece of news concerns the 1998 ports some sad news, ACM Kanellakis Theory and Practice some news that is Award. This was given to Randal E. Bry- cause for celebration ant, Edmund M. Clarke, Jr., E. Allen but makes us recall Emerson, and Kenneth L. McMillan for some sad events of the their invention of “symbolic model check- recent past, and some ing,” a method of formally checking sys- news about changes tem designs that is widely used in the in the life of a friend computer hardware industry and is be- and colleague that are ginning to show significant promise also cause for both reflec- in software verification and other areas. Tom Dean tion and celebration. This award was established in memory of our friend and colleague Paris Kanel- First, some very sad news. Markus Meis- lakis, whose tragic death in late 1995 cut ter, in his second year of Ph.D. study in short a distinguished research career. the department, died in a Christmas Eve Paris and his family are frequently in our automobile accident in British Columbia. thoughts and we often hear from their Markus was well loved in our commu- many friends, students, and colleagues nity. In addition to his research in graph- from around Brown and throughout the ics, Markus was prankster and pros- world. elytizer for classical music, mathematics for its own sake, and all things concern- Finally, on April 22 we had a party for ing Japanese culture. His whimsical side Peter Wegner, who will be retiring at the was obvious in his interest in anime vid- end of this semester. We don’t expect to eos, particularly those including the char- see any less of Peter in his new role as acter Totoro, and in his fascination with emeritus professor, and indeed his retire- all sorts of hardware, including sports ment party was cause for a celebration of cars well past their prime but nonethe- his scholarly achievements and his many less beautiful. More than anything else, contributions to the department. At the

conduit! 16 Enjoying his retirement party at the Faculty Club. Clockwise from bottom left: Peter with President Gee, with long-time friends from the Sociology Department Marilyn and Dietrich Rueschmeyer, and with CS chairmen past and present

party President Gee thanked Peter for his long service to Brown and announced the news that Peter would receive the Medal of Honor for Arts and Sciences from the President of the Austrian Federation at a ceremony this summer. President Gee re- marked on some of the extraordinary events in Peter’s life, from his escape from Austria as a child on the “Kindertransport” to the award of the Austrian Medal of Honor almost exactly 60 years to the day later. There is certainly no reason for sadness that Peter has passed this milestone in his career; quite to the contrary, we look forward to following his scholarly achievements and comic shenanigans for decades to come. The sentiment on the brass plaque affixed to the classic spindle-back Brown chair that he was given at the party read, “To an unrepentant punster, inde- fatigable scholar, and generous friend.”

located at 180 George St. This lab will be BROWN ENTERS THE CAVE ERA available as a research, computing, and education facility and is expected to fos- In the fall of 1997, six Brown depart- ter scientific investigation and graduate ments—Applied Math, Chemistry, Cogni- and undergraduate instruction in a vari- tive Science, Computer Science, Geology, ety of fields, including chemistry, geology, and Physics—were awarded a joint Na- cognitive and linguistic sciences, phys- tional Science Foundation MRI (Major ics, computer science and applied mathe- Research Infrastructure) grant to devel- matics. The lab contains a high-perform- op a university-wide computing facility ance parallel computer, the IBM SP, and for collaborative research and educa- a cave. tion. The result is a new supercomputing and immersive virtual reality lab—the What is a Cave? Technology Center for Advanced Sci- Brown’s cave is an eight-foot cubicle in Rosemary Simpson entific Computation and Visualization— which high-resolution stereo graphics are

conduit! 17 Illustration of CAVE™ in use projected onto three walls and the ﬂoor to wide with applications ranging from sci- create an immersive virtual reality expe- entiﬁc visualization, industrial research, rience. High-end workstations generate and training, to education, theatre, and the 3D virtual world and create the the arts. (See http://www.cs.brown.edu/ sounds of the environment. Special hard- research/graphics/research/vr.html for cave-related resources and applications.)

“...it is easier to understand Brown’s Technology Center for how something works or is Advanced Scientific Computa- put together if you can tion and Visualization The Technology Center resources include hold it in your hand or walk an IBM RS/6000 SP parallel computer and the cave. The IBM computer will be around inside it” used to simulate complex processes as diverse as the movement of the earth’s ware and software keep track of the posi- mantle, subatomic particle interactions, tions and movements of a person entering and the function of the human heart. It that virtual environment, changing the will also drive the projectors that display images in the cave in a way that allows those images in the cave. For example, a the visitor to feel immersed in the virtual researcher studying a simulation of air space. flow around a space shuttle can walk around a model of the shuttle floating Our cave is based on the original CAVE™ within the cave, even feeling compelled to developed at the Electronic Visualization duck while walking under it. The re- Lab at the University of Illinois, Chicago, searcher could see patterns of airflow as part of long-term research on tools and around the shuttle by positioning and applications for immersive virtual reality. moving colored streamers in the flow It was first demonstrated at the SIG- much like tails on a kite. “We anticipate GRAPH ’92 conference. Since then, doz- that the cave will become a powerful tool ens of caves have been installed world- for facilitating scientific insight,” said

conduit! 18 Samuel Fulcomer, Center director. “This Auditorium of MacMillan Hall, 167 Thay- is based on the idea that it is easier to un- er St. and will feature Brown President derstand how something works or is put E. Gordon Gee; Andries van Dam, the together if you can hold it in your hand or Thomas J. Watson Jr. University Profes- walk around inside it.” sor of Technology and Education and Pro- fessor of Computer Science; George People interacting with the cave’s graph- Karniadakis, Professor of Applied Math- ics wear tracking devices to monitor their ematics; and Paul M. Horn, IBM Senior movements, lightweight stereo eyewear Vice President of Research. After the cer- to see objects in 3D, and slippers to pro- emony, there will be demos at the 180 tect the delicate polymer screens lining George Street facility, among them a the cubicle. The projected virtual space Mars topology demo by the Geology De- appears seamless, without any intersec- partment and 3D modeling and an archi- tions between ﬂoor and walls. tectural walkthrough of a virtual environment by the Computer Science De- partment. Computer Science Application Possibilities Like all the departments collaborating in the cave facility, Computer Science is developing research and educational applications that would either be impossible or much less effective without it. Some of these areas include:

Scientific visualization: In conjunc- tion with NASA, the Graphics Group has been doing computational fluid dynamics studies that will benefit from the cave’s environment. New studies of bloodflows are also being developed. Professor David Laidlaw is studying biomedical areas where he points out that graphically- based studies of deep structure and tis- sue in areas like the brain, the hand and CAVE™ showing projection system the spine will enormously benefit from immersive 3D visualization. As viewers negotiate a virtual environment, objects may appear to hang in Concept visualization: Professor Eli space in front of them, tangible and with- Upfal states that both research and in reach. This is an effect of the stereo teaching of theoretical areas would bene- glasses, which create slightly offset imag- fit from access to an immersive 3D visual- es for the left and right eyes. The cave us- ization facility. For example, network er may also wear special gloves that send analysis studies involve multiple dimen- the computer information about hand lo- sions and he feels that the development cation and thus let him or her interact of intuition about these processes would with the virtual environment using ges- be greatly accelerated by such a facility. tures. A viewer could change the lighting and artwork in a virtual gallery, precisely Algorithm visualization: Professor position an artificial valve in a virtual Roberto Tamassia currently uses exten- heart, or fly over a virtual Martian land- sive visualization techniques for algo- scape. rithm teaching and research and feels that 3D immersive techniques would be Dedication Ceremony highly beneficial. He pointed out, howev- The dedication ceremony will take place er, that the small size of the facility would at 2pm on May 26, 1999, in the C.V. Starr severely limit its usefulness in teaching.

conduit! 19 Behavior simulation: Professor Nancy work solely in a desktop keyboard-screen- Pollard and her students have been mouse mode. New approaches that incor- studying interactive teaching situations. porate new gestural, vocal, and visual vo- One project, called Coach, is a 2D/3D cabularies are needed to work in these electronic playbook for football. Coaches environments. can develop and modify football plays and Research and educational opportunities players can view animations of those exist in all of these areas, as well as po- plays within a 3D environment (e.g. from tentially in other domains such as sys- a ﬁrst-person point of view). A second tems analysis. For further information on project explores animation of dance mo- research, support, and demos, contact: tions for archival and teaching purposes. Director Sam Fulcomer at sgf@cfm. brown.edu: User interface research: Graphics Group members Andy Forsberg and Joe LaViola have been investigating multi- modal user interfaces needed in computing environments where users no longer

OUR MOST RECENT PhDs

NAME ADVISOR THESIS POSITION

Mitch Stan Zdonik “Building Query Opti- Assistant professor at Cherniack mizers with Combina- Brandeis tors” Hagit Shatkay Leslie “Learning Models for Postdoctoral fellow at Kaelbling Robot Navigation” the National Center for Biotechnology Information. She is developing methods for faster and more accurate biomedical information retrieval and on applying machine-learning-type methods in bioinfor- matics Laurent Michel Pascal Van “Localizer: A Modeling Ilog S.A., Belgium Hentenryck Language for Local Search” T.M. Murali Jeffrey “Efﬁcient Hidden-Sur- CS post-doc at Stan- Vitter face Removal in Theory ford working in the and in Practice” robotics and graphics labs with Professors Leo Guibas and Jean- Claude Latombe, among others

conduit! 20 Hagit and Eadoh complete a construction project

Laurent Michel and poultry after his defense and cele- brating with his wife, Valerie

Mitch Cherniack

and as Heidi’s colleague is a government CHARNIAK UNPLUGGED employee, if the government did shut down, he could not go (or at least could A few conduit!s ago I wrote an article not be reimbursed for travel expenses) entitled “Bad Trips,” about problems that because technically he would be unem- have come up in faculty’s trips to give ployed. Thus Heidi had to plan on going talks, etc. The problems were mostly and since the government eventually did standard, from bad travel connections to not close down, they both were there. bad weather. However, I just heard a Having both authors of a paper show up new class of problem from one of my at this conference seems to have been graduate students, Heidi Fox. She and a pretty unusual. Actually, just getting one colleague were invited to give a talk at a author to show up was not that common conference on some work they had done either. For the four or so papers in the jointly. They agreed that her colleague session, only three authors showed: would give the talk, and thus Heidi was Heidi, her colleague, and one other Eugene Charniak not planning on going at all. However, author. On the other hand, perhaps the this was the end of last year, when there missing authors knew something that was a threat of a government shutdown, Heidi did not—showing up was clearly

conduit! 21 optional, since the speakers outnumbered sign (hastily improvised by attaching the rest of the audience. cardboard to a fly swatter) that adver- To understand the following bit of depart- tised the conference. The sign said, with mental goings-on, you need to know one an admirable economy of words, “JAVA”. But while the speakers knew how to piece of university/department trivia. interpret this, the average passerby, The university distinguishes between needless to say, did not, and this at- very small purchases and bigger ones. tracted even more attention than walking Very small are handled from petty cash. down supermarket aisles with $130 While this is convenient for small pur- worth of cookies (“Where’s the free coffee, chases, the amount of paperwork per dol- lady?”) The fly swatter is now retired to lar is clearly higher than for larger ones. the CS archive. Thus when the department ran out of cookies last fall (they are served before Avi Silberschatz of operating systems departmental talks, along with tea and fame was a speaker in the department’s coffee), Jennet Kirschenbaum figured distinguished lecture series this year. In that we should stock up for a while and his talk he mentioned a common OS tech- had a purchase order made out to the nique of keeping twice as much storage tune of $125. Then she and Mary as one needs for an IO-bound task. One Andrade walked through a local grocery then reads into half of the storage while store (the only one that takes Brown pur- one is reading out of the other half. chase orders) with calculator, trying to When the input half is all filled up, one buy exactly $125 worth of cookies. This just switches halves. In his talk Silber- schatz mentioned that Stan Zdonik told “...the department goes him of someone who did this with dishes. I immediately knew what he was talking through130 dollars’ about because many years ago I decided that it would be a great idea to have two worth of cookies in six to dishwashers. One puts dishes into one seven weeks” when they are dirty, then when it is full one runs it. In the meantime one is removing dishes from the other dish- turned out to be a shopping cart full, washer. The point, of course, is that one which, as you might expect, attracted a never has the task of removing dishes to lot of attention (“Can I come to the a cabinet. After the talk Stan told me party?”) The bill was $130. Jennet esti- that the person in question is Tom, one of mates that we use about $10 of cookies the Click and Clack brothers. In re- per event. Figuring about two events per searching this article I called Stan to ask week, the department goes through 130 whether Tom was Click or Clack. Stan dollars’ worth of cookies in six to seven said that this is a deep philosophical weeks. In the meantime, Jennet stores question about which there is a great the cookies “on the 5th floor—locked up.” deal of controversy. He did give me a bit of fast-breaking news however; namely, On a related topic, as mentioned else- that MIT, after having U.N. Secretary- where in this conduit!, our fall IPP sym- General Kofi Annan as its graduation posium was on the Java programming speaker two years ago and President language. Since parking is always tight Clinton last year, is going to have Click around Brown, Suzi Howe had spaces and Clack this year (they are both MIT reserved for the speakers near our build- alums). Remember you heard it here ing. However, in order to direct them to first! the spaces, and in order to make sure that they bunched up the cars so they I learned about an even more ambitious could all fit in, Mary Andrade and Dawn home-improvement scheme a month or Nicholaus took turns patrolling outside two later when I attended the Neural In- the building to intercept the speakers. As formation Processing conference to give they had no idea what the speakers an invited talk. I noticed that Mike Moz- looked like and vice versa, they carried a er, who graduated from Brown about

conduit! 22 1980 or so and worked with me on an AI perature to save on fuel, and even turn project during his senior year, was also on lights when he is about to move to a speaking, and we got together to bring different room (it notes motion in the each other up to date on what was new in room and has learned that at the particu- our lives. Mike is now a professor of com- lar hour in question Mike usually leaves puter science at the University of Colo- that room to go to a second room). His house, as you might imagine, has been rado, and one of his projects is wiring up written up many times (both academical- his entire house so that his home comput- ly and in the popular press), and has its er knows things own Web page:....www.cs.colorado.edu/ like (a) whether he ~mozer/adaptive-house. One interesting is in the house and result of this project is that Mike’s water if so, in which heater is the property of the University of room, (b) what Colorado. lights he has on, (c) Finally, on the day of one of last winter’s what he has the larger snowstorms, Dawn Nicholaus heat set to, etc. As brought her five-year-old daughter Julia Mike is particular- to work with her for a few hours. When I ly interested in ap- emerged from the elevator on the fourth plications of mach- floor, Julia was sitting there and greeted ine learning (which me with a cheerful “Hi, bro!” Not expect- is why we were ing a greeting at that point, and certainly both at this confer- not expecting that particular honorific, I ence), his basic managed only a weak “Hi” back. Later idea was to get the that day Dawn and I discussed whether a machine to learn middle-aged white professor could be a his habits and pref- “bro,” and eventually decided that if Julia Dawn and Julia erences so that it thought so, then yes. For the next few could do things like turn off lights that days the administrative staff called me were not in use, raise and lower the tem- ‘‘bro.”

companies, and served on the advisory “ ANDYFEST ” boards of many small and large companies, including Microsoft Research. He To celebrate Andy van Dam’s has also mentored a steady stream of un- 60th birthday, a symposium dergraduates, many of them current or will be held May 27-28 entitled former chairs of the top-ranked depart- “The Computer, the Academy, ments in the country. and the World.” Among his awards are the ACM SIG- GRAPH Steven A. Coons Award (1991), Andy was awarded the second the ACM Karl V. Karlstrom Outstanding US PhD in Computer Science Educator Award (1994), and the IEEE in 1966. Since then, through James H. Mulligan, Jr. Education Medal his deep commitment to educa- (1999). In 1994 he also became an IEEE tion and his boundless energy, Fellow and an ACM Fellow. He has hon- his impact on the emerging orary PhDs from Darmstadt Technical world of computers has been University in Germany (1994) and Swar- enormous. He has broken new ground in thmore College (1996). In 1996 he was the use of computers in education, educat- inducted into the National Academy of ed a whole generation of experts in com- Engineering. puter graphics (not to mention co- Speakers at the two-day event will be founding ACM SIGGRAPH, whose annual Ronen Barzel, who joined Pixar in 1993 conference draws more than 30,000 at- to work on Toy Story and has since tendees), helped with countless startup worked on R&D of modeling, lighting and

conduit! 23 animation tools; Ingrid Carlbom (’80), the inventions of the now ubiquitous Head of the Visual Communications Re- overlapping-window interface and mod- search Department in the Multimedia ern object-oriented programming and Communications Research Laboratory at whose numerous honors include the Bell Labs; Steven J. DeRose (’89), co- ACM Software Systems Award; Ed founder of Electronic Book Technologies Lazowska (’72), Professor and Chair of and now Chief Scientist at Inso Electronic the Department of Computer Science Publishing Systems; Henry Fuchs,Fe- and Engineering at the University of derico Gil Professor of Computer Washington; Raj Reddy, Dean of the Science and Adjunct Professor of School of Computer Science at Carnegie Radiation Oncology at the Univer- Mellon University and the Herbert A. sity of North Carolina and winner of Simon University Professor of Computer ACM SIGGRAPH’s 1992 Computer Science and Robotics, who received the Graphics Achievement Award; Dr. ACM Turing Award in 1995; and David Alan Kay, Disney Fellow and Vice Salesin (’83), Senior Researcher at President of Research and Develop- Microsoft Research and an Associate ment, The Walt Disney Company, Professor in the Department of Comput- who is best known for the idea of er Science and Engineering at the Uni- personal computing, the conception versity of Washington. of the intimate laptop computer, and http://www.cs.brown.edu/~andyfest

conduit!conduit! A publication of The ComputerA publication Science ofDepartment The ComputerBrown Science University Department Brown University❦ Inquiries❦ to: conduit! Inquiries to: conduit! Department of Computer Science DepartmentBox 1910, of Brown Computer University Science BoxProvidence, 1910, Brown RI University 02912 Providence,FAX: 401-863-7657 RI 02912 PHONE:FAX: 401-863-7657 401-863-7610 PHONE: 401-863-7610 EMAIL: [email protected] Suzi Howe Katrina Avery Jeff Coady WWW:EMAIL: http://www.cs.brown.edu/ [email protected] Suzi Howe Katrina Avery Jeff Coady WWW: http://www.cs.brown.edu/ Editor-in-Chief Editor Technical Support publications/conduit/ Editor Technical Support publications/conduit/ Editor-in-Chief

Department of Computer Science Brown University NON-PROFIT Box 1910, Providence, RI 02912, USA U.S. Postage PAID Providence, RI Permit #202

conduit!

Printed on recyled paper Address changes welcomed

conduit! 24