Parent


An Electronic Journal Browser Implemented in the World Wide Web

Marc E. Salomon, B.A.

David C. Martin, M.S.


The University of California, San Francisco
Center for Knowledge Management
Innovative Software Systems Group
San Francisco, California 94143-0840


Abstract 

The networked delivery of medical journal content along with innovative presentation of associated abstracting and indexing data presents new issues for the evolving digital library. This paper examines some of the strategies deployed towards the development of a World Wide Web client for the Red Sage project in AT&T's RightPages electronic journal browsing project using freely redistributable networked information discovery and retrieval tools available in the public domain. Work focused on the development of a suite of HTTPD-server-side Perl scripts to navigate the file system data tree, bibliographic information acquisition schemes, and WAIS indexes. Scanned TIFF page images as well as OCR data are provided by AT&T. Abstract and indexing information is retrieved from NLM's (National Library of Medicine) MEDLINE index of health sciences articles. The scanned text and the MEDLINE records are indexed by the freeWAIS full-text indexer both on a collection-wide basis as well as per-journal. In addition, the NLM authorized Medical Subject Headings are extracted from the MEDLINE records and separately indexed to provide an interface to the search engine. The resulting search scheme provides both manual entered term searching as well as authorized term searching. Search results provide extended bibliographic data along with an interface to search the authorized terms for each article and a hypertext link directly to the scanned text images for that article. The NCSA HTTP daemon was modified to include code from the UNIX file(1) in order to determine the content type from the actual data in the file instead of relying on the file extension, which is not descriptive as provided by AT&T. A graphical WWW client capable of displaying TIFF images is required to use the system.

1. Introduction

With the rapid increase in computer processor speed, data storage capacity and network band-width, the academic research library is poised to emerge from its formerly limited role as centralized repository for print material to move toward a more ubiquitous presence as a provider of knowledge management tools [Lucier 90]. The changes that libraries face in moving from a paper collection towards a center for information dispersal are similar to those faced by publishers, whose entire modus operandi is challenged by the migration from paper-based information dissemination in favor of electronic content delivery. Towards working through some if these novel issues, the University of California, San Francisco, in collaboration with AT&T and publisher Springer-Verlag, have implemented the Red Sage on-line journal delivery project.

1.1. Red Sage Project

Based on AT&T's RightPages client-server journal browsing system, the Red Sage collection comprises approximately 60 journal titles. The publishers supply AT&T with the paper instances of each journal issue or scanned TIFF images. AT&T produces a set of variable magnification TIFF image scans of each page, a set of bounding box coordinates for critical elements on the page, as well as the output of an OCR for each page in the form of horizontal and vertical positioning information for the box that bounds the element. The browser is a MOTIF graphical user interface client which communicates with a server using the Internet TCP/IP network protocol suite. The server manages the data store of TIFF images and provides an interface to the Slimmer full-text database.

1.2. WWW Red Sage Server

Since the influx of content is regular and constant, one priority in the development of the World Wide Web (WWW) Red Sage server was to keep the maintenance of HyperText Markup Language (HTML) pages to a minimum, thus the reliance on the HTML-generating server-side Common Gateway Interface (CGI) [McCool 94a] Perl scripts to ensure the highest level of confidence in the results. At each level in the stack, from the level of journals in the collection down through the journal-specific levels such as year, volume issue and article, the server-side CGI scripts examine the on-line available content and produce HTML that reflects actual, not theoretical content thereby all but eliminating faulty links.

Another goal was to reproduce the functionality of the AT&T product completely. As far as browsing, we have provided the user with equivalent functionality for navigating journals, years, volumes, issues, pages, articles, stacks and figures. The freeWAIS-based full-text search system performs comparably to the Red Sage Slimmer database, while the MeSH search capability performs favorably with the AT&T search engine while providing enriched functionality to essential to our patron community of biomedical researchers. We have delayed implementation of the alerting feature of the AT&T product, but have succeeded in integrating the content into a WWW client for the University of California's MELVYL bibliographic system.

1.3. WAIS Indexing

We employ the WAIS full-text indexer and search engine to provide term searching for the system. The error-prone OCR text provided by AT&T is stripped of its positioning information and combined with a MEDLINE record if available. Not all journals are indexed by the National Library of Medicine (NLM), so coverage is lacking in some cases. Each HTML page generated at all levels contains a hypertext link to the search page. We also build WAIS indexes from the combined MEDLINE record and OCR text on a per- journal basis, and the user is allowed to specify a set of journals on which to search as well as perform a full-text search across the entire collection. When the MEDLINE record is available, we extract the MeSH terms that are a vocabulary of authorized terms sanctioned by the NLM and build a WAIS index of just those terms, weighting terms deemed particularly important by the indexers. In order for the user to enter the authorized term, we create a set of webs that allow access to the MeSH term for searching through an alphabetic approach or by exploiting (to a limited degree) the hierarchical nature of the Medical Subject Headings. In each case, the final selection is used as input to the searching routines and the results of the MeSH WAIS search are returned as any other WAIS search result set.

1.4. MELVYL/Red Sage Client

One goal of the Center for Knowledge Management is to create a seamless interface to the various Internet resources of interest to the biomedical academic research community. To that end, we have sought to integrate existing resources so that the distinctions between this prototype project and well-tested and used resources, such as the University of California System's Division of Library Automation's (DLA) MELVYL bibliographic indexing system are minimal. We have developed a prototype system allows the user to enter a bibliographic search to the DLA's MEDLINE database through an HTML form, with the result set presented to the user as a set of HTML enriched MEDLINE records that include hypertext links to the TIFF page images of any results from the search that have content in the Red Sage collection.

1.5. Further Research

As any new technology is first deployed, at least as many questions are posed as are answered by the research and implementation. In this case, the limitations of the current versions of HTTP (HyperText Transfer Protocol) and HTML, designed for near-interactive information interchange between researchers in rapidly evolving fields have begun to manifest themselves as that technology is applied as a production information presentation protocol. In addition, in reexamining the RightPages system and the limitations of AT&T's product (if any) present opportunities for further research and development. The remainder of this paper shall focus on an in-depth discussion of the issues outlined throughout this overview.

2. Red Sage Project

The University of California San Francisco Library and Center for Knowledge Management (CKM) is taking an important step in the creation of a knowledge management environment. Through the Red Sage project, a collaborative effort founded by CKM, AT&T Bell Laboratories and Springer-Verlag, the collaborators are exploring the electronic distribution of journals directly to individual physicians and faculty computer desktops.[Lucier 92] With RightPages software developed by AT&T Bell Laboratories, researchers, clinicians, and students will be able to search, read and print the full-page text, including graphics and photographs, from an initial collection of journals published by Springer-Verlag, John Wiley & Sons, the Massachusetts Medical Society and a number of other prominent publishers.

2.1. Participants

With this project the participating organizations will investigate and begin to understand the technical, legal, economic, business, intellectual property rights and human factors issues surrounding the creation, distribution and use of scientific and medical information in a networked environment. As the first in a series of CKM originated projects, Red Sage will begin to examine the issues that are critical to the creation of the digital library of the future. Initially the project will run on a local area network with a central document database server and multiple user stations. Each member of the Red Sage project is contributing resources at its own expense. Springer-Verlag and the other publishers are providing electronic content. AT&T's Bell Laboratory is providing content preparation, quality control, and the server software and the initial client application. CKM is providing the test population, delivery infrastructure, evaluation and subsequent client development.

2.1.1. Publishers

Each publisher is initially providing monochrome page scans at 300 DPI, separate gray-scale figures (e.g. photographs, diagrams and tables) and structured1 header information (e.g. author, title, etc...). Additional content formats will be generated by the publishers in the second phase, including PostScript, Standard Generalized Markup Language (SGML - ISO 8879) tagged header, reference and body text, and potentially original data sets and other research results.

2.1.2. AT&T Bell Laboratory

The Bell Laboratory of AT&T has developed, and will continue to refine, the RightPages server and client software. The client software is a graphical user-interface (GUI) application for UNIX, Macintosh and PC computers under MOTIF, the Macintosh Finder and Microsoft Windows GUI environments. AT&T is also providing instrumentation and performance monitoring tools for the RightPages server, as well as determining the content delivery form(s) that publishers will be providing throughout the project, especially as publishers are able to provide SGML text for article body and reference sections.

2.1.3. The University of California San Francisco

The Library & Center for Knowledge Management (CKM) manages the central server repository and its content, supports the local user community and determines infrastructure (e.g. server, network, staffing, etc...) requirements and scalability. CKM is providing multiple system access terminals throughout the campus centers as well as within the Library. In addition, software engineers are developing next-generation client interfaces and investigating links between the Red Sage content and other data sources (e.g. bibliographic, informal communication, scientific databases, etc...). Members of the CKM staff will also evaluating the implementation and effectiveness of the project. These evaluations consist of pre- and post- deployment interviews with a targeted user community, analysis of usage data generated by client-server interactions (including hard copy requests), and continuous monitoring of the system's performance.

2.2. Technical Overview

The Red Sage Project has been assembled from a variety of sources. The founding partners, CKM, AT&T and Springer-Verlag envisioned the electronic distribution of published literature via computers and networking. The RightPages software from AT&T, developed at the Bell Laboratory [Story 92], provided the initial technology: a client/server model with a central server repository. In 1992, the three partners agreed to proceed forward; AT&T providing the software, Springer-Verlag the content, and CKM providing the test-bed. In this section the technical subsystems are examined and explained; they are: system requirements and implementation; content preparation, delivery and management; and client/server architecture.

2.2.1. System Requirements and Implementation

The Red Sage Project incorporates a client/server architecture with three main aspects to the system infrastructure: server, network and client. The server provides the data store for the journal content, computational support for the server processes that build, maintain and search content on behalf of client requests, and host processing for a number of directly supported access terminals. The network connects the client to the server via the de facto Internet standard TCP/IP protocol, either on a local-area or wide-area topology. The client application uncompresses images from the server, displays content with a graphical user interface and supports printing to local printers.[Story 92]

2.2.1.1. Networking
The RightPages client and server communicate via the Internet de facto standard TCP/IP protocol. This protocol is supported across both the local-area and wide-area networks that connect the distributed facilities of the University of California San Francisco and its affiliates. The RightPages server system is located in the Library on the main Parnassus Heights campus; several access terminals are available in the Library and utilize the installed 10BASE-T Ethernet network. There are also access terminals, provided by the Library and Center for Knowledge Management, located at San Francisco General Hospital, Mt. Zion Hospital and at the Veteran's Administration Hospital. Each of these other facilities is connected to the Parnaussus Heights network via 1.54 M-bit/sec (T1) link and supports Ethernet locally.

3. WWW Red Sage Server Implementation

The statelessness of the WWW HTTP presents some interesting problems for the server-side script developer. Each element of state must be preserved within the URL which is passed to the CGI script through the appropriate CGI environment variable. For the purposes of this accessing a page in a Red Sage journal article, the elements of state are: At each level, we are able to determine all appropriate information from that point in content space from the above elements. In the case where we arrive at a page image through navigation by the HTML generated by the server-side scripts and final translation of a click on an table of contents entry, this is trivial. The scheme also functions properly if the above elements are provided through a different means, such as an external bibliographic system. In the case of searching, we use the five elements of state described above to parse the issue structure file provided by AT&T to produce a pointer to the first page of an article.

3.1. Top Level Navigation

When the patron opens the URL that points to the top-level of the Red Sage collection, the CGI script examines all entries at the highest level of content. A list of all journals in the collection is presented as an HTML page containing greyscale TIFF images if the latest issue for each title. The server- side software queries the data store for the current status of the contents at that level so faulty links are avoided. We can get this information either by examining the file system directly or by issuing a query to the Red Sage RightPages server daemon. The greyscale icons for the most current contents for each subsequent level are organized and presented in LIFO order. Thus, the patron can descend down from the collection of journals, through years, volumes and issues, finally reaching the first page of the text after selecting an article entry in the table of contents. At each level, the CGI-generated HTML page contains hypertext links to the parent page as well as to the appropriate child pages, the top-level page and a search page.

3.2. Table Of Contents

After navigating to an article in an issue, the user is presented with an HTML page containing a TIFF image of the table of contents page or pages. At the present time, AT&T adds value in the form of a file for each issue which contains position and bounding box coordinates for each item in the table of contents. The HTML tag is used to pass the coordinates of the user's mouse click on the Table of Contents to the server, which returns an HTML page with the appropriate first page image and navigation buttons for the selected article. Along with each issue AT&T provides data files which describe the position of each page relative to its article as well as pointers to any high-quality figure images associated with each article.[O'Gorman 92]

3.3. Page Delivery

Once we reach the terminal level; that of content provision, the server-side Perl [Wall 90] scripts select the appropriate full-text TIFF page image for the selected resolution and builds an HTML page around it. Note that the TIFF image format is not currently supported as an inline option for NCSA's Mosaic World Wide Web browser. Cheong S. Ang, a software engineer at the CKM has incorporated the public domain TIFF library code into the browser so that the images are presented as inline images on the page. At the head of each page are four sets of buttons; one set for page-wise navigation, one for article- wise navigation, a third for stacks navigation and the fourth for miscellaneous options. The footer of the page contains a link to any high-quality images, currently JPEG images of the figures associated with each article. Since these pages are always generated on-the-fly, we are assured of the validity of each link. If the software can't guarantee the validity of a particular link, whether for lack of content or failure to convert an external bibliographic reference to content (content anomaly) it is either grayed out of replaced with the next-best choice. The first and last of each set are almost always active while context determines the state of next and previous. The sets of buttons are described below.

3.4. Content Navigation Buttons

3.4.1. Page-Wise Navigation

These buttons provide the ability to navigate within the article. First and last page of the current article are always active, while next and previous page buttons are enabled if appropriate. We query the positioning files provided by AT&T in the data store to determine the position of this page at the article and issue level. Since we pre-resolve all hypertext links checking for the existence of the actual TIFF contents the system always returns an valid pointer. The hypertext link generated is thus guaranteed to be as valid as the data store is accessible.

3.4.2. Article-Wise Navigation

Similar to the page navigation functions, the article buttons allow navigation within the issue according to the page positioning data provided by AT&T. The article buttons function as their page-wise analogs.

3.4.3. Stacks Navigation

At each level, we generate a set of stack pointers that allow navigation between levels without having to reload images in the skipped levels. One drawback that the current HTTP/HTML client-server architecture presents is the inability to force-cache images that can require constant reloading. With some sixty (60) iconic images on the top level of the collection, the overhead of that many HTTP connections rapidly approaches that of the data transfer itself. If we have learned anything about the limitations of HTML and HTTP as implemented now from this project, it is the need for both always/never caching hints to the browsers to and the incorporation of an MGET facility in HTTP. At present, we provide direct access to the top level and the volume level from each page image.

3.4.4. Miscellaneous Navigation

At the level of content delivery, Navigational options include several links to additional features of the system. The user can select a magnification level for page images using a zoom toggle button. This choice stays in effect for the duration of the session. A link to the script that generates the WAIS search page is included in this section, as is a link back to the table of contents for the current issue. A link is included to an HTML enriched presentation of the MEDLINE record for the article, if available. In the case that the MEDLINE record is available, the system attempts to build a web of enriched MEDLINE abstract pages for each article in the issue which can serve as an annotated table of contents.

3.5. Figures

AT&T currently provides high-quality JPEG scans of any figures that might accompany an article. The assignment of images to articles is controlled by the group description files associated with each issue. If the CGI scripts detect any figures for an article, links are placed at the bottom of the page and the browser calls the appropriate visualization program to present the image to the user.

3.6. Searching

In order to reproduce the functionality of the AT&T system, we chose the WAIS index and search engine to build an inverted index of the full-text and MEDLINE information. When a user clicks on that hypertext link, a CGI script builds a search page dynamically that: After the search is complete, an HTML page is generated that allows the user to view the results as a set of formatted MEDLINE records that include bibliographic information, the full abstract, if available and the ability to search the MeSH index using the authorized MeSH term. From either the raw search results or the enriched MEDLINE record, the user can jump directly to the full-page TIFF image by a hypertext link.

4. WAIS indexing

In addition to the OCR of the scanned text image supplied by AT&T , we query the MELVYL MEDLINE index of biomedical journals for extended bibliographic data on each article. In order to preserve the integrity of the AT&T supplied content, we create a mirror tree of MEDLINE records on traditional magnetic disk that corresponds to the TIFF data on MO storage. From this parallel tree, we generate the WAIS indices used in the searching features of the system.[Kahle 89]

4.1. Full Text

AT&T provides an OCR of the scanned text page as a set of coordinates describing the position and dimensions a bounding box and the corresponding text. This information is not useful for reconstructing the logical content for such applications as presentation through an ASCII WWW client, but can be used for a full-text search. We remove the coordinate information from the file, and combine it with the NLM MEDLINE index record. A WAIS index is then built of all article's MEDLINE record and scanned text file, with the WAIS Headline field in the Doc ID overloaded to present a URL as the headline when searched.

4.2. Fielded Search

We have also deployed the WAIS indexing software to construct a brute-force relational database of sorts. The MEDLINE record for each article is compiled by the NLM and provides a set of authorized terms from the hierarchical Medical Subject Headings (MeSH) system. We create a WAIS index for the whole collection as well as for each journal of the MeSH terms in the constituent MEDLINE records. Since the MeSH is an controlled vocabulary, we create a web of HTML that allows the used to select a search term either alphabetically or through the hierarchical sub-headings. This is one of the pre-cooked aspects of the system in that we must perform a download of MEDLINE records from MELVYL and then process the entire result set. Downloading the records from MELVYL takes several hours, while the Perl script that generates the MeSH web takes approximately 20 minutes. The searches are quite speedy, less than 1 second, due to the small size of the MeSH terms.

5. WWW MELVYL/Red Sage Client

One of the appealing features of the World Wide Web is the ability to effectively obscure the specific details of the precise network tasks performed upon the activation of a hypertext link. For the patron of the biomedical research library the details of the network mechanics involved in linking any number of related information stores for the maximum value added are irrelevant. Towards that seamless knowledge management environment, we have developed an HTML interface to the DLA's MELVYL index of the NLM's MEDLINE annotated index of biomedical journal articles. Using a forms-based search page, a patron enters a search that starts a telnet(1) session to the MELVYL server. A connection is made to the MEDLINE database, and a search is issued. Upon successful completion of the search, the server processes the raw MEDLINE record, presenting the author, title, journal name, volume and issue as well as the abstract. If the server can determine that the article returned by the search is resident in the Red Sage data store, it will also include a hypertext link that interfaces to the page-level content delivery CGI script. In this manner, the we have effectively linked the two distinct yet overlapping bibliographic systems.

6. Modified HTTP Server

We have been using the National Center for Supercomputing Applications httpd Hypertext server for delivering this content over the World Wide Web.[McCool 94] A fine piece of software, we found it necessary to perform a few minor enhancements in order to serve the particular content as provided by AT&T. As a convention, most all World Wide Web clients and servers overload the filename with an extension in order to determine the type and the proper presentation method. This legacy of the archaic MS- DOS operating system was wholly inappropriate to huge data stores developed under environments far from the desktop personal computer. The Red Sage TIFF images do contain the letters 'tiff' in the filename, but they are not at the end of the filename after a period. In order to avoid either renaming gigabytes of TIFF to satisfy the filename overload convention to get the MIME content-type: set correctly or establishing a set of symbolic links with overloaded names to the original content, we chose to modify the server. There exists a standard UNIX utility, file(1) which checks the magic(5) number, a unique bit sequence that identifies the type of a file, and sets the MIME content-type: field accordingly. We integrated this code into the NCSA httpd server so that it bases the value of the content-type: field on the actual contents of the data file instead of an uncertain file extension.

7. Further Research

This project has been a prototype development effort, and in conducting the research required to produce this prototype, the opportunity for further enhancements have surfaced. Since we are working from a commercial product supplied by AT&T,. the needs of the academic research library diverge from the goals held by the developers of the original system. Working towards a knowledge management environment, the mission here at the CKM is to integrate various networked information resources into a seamless, coherent user interface. The ease of use of the World Wide Web system forces the complexities behind the scenes so that the details of the network magic are hidden from the user. Since integrating a large MOTIF RightPages client into Mosaic or any other WWW browser is impractical, we need to build upon the work done by the CERN, NCSA and others to build systems capable of delivering content to patrons with a minimum of effort and potential confusion. A discussion of a few of the more interesting points that the aforementioned development effort produced follow.

7.1. Alerting

The AT&T system provides for an alerting system based on user-established profiles. We would duplicate the functionality of the Red Sage system by implementing a cron(1) job that would collect the log files for newly loaded content and generate a hit list for each users' profile entries. based on a newly added content WAIS index created for this purpose. Moving further on this model, we would like to design a more generalized alerting system that would query MELVYL MEDLINE for newly added content, provide an alert list, either through electronic mail or a HTML page with links to the page images of any overlaps in the MEDLINE result set that were also resident in the Red Sage content.

7.2. Automatic Bibliographic Updates

At the current time, we perform a bulk download of MEDLINE records from the DLA MELVYL system with no regularity. We would like to establish a system where the aforementioned content load log files triggered a telnet(1) session that connects to MELVYL and downloads the MEDLINE record. In addition, this would tie into the alerting subsystem as described above. We are considering employing the auto update feature of MELVYL to initiate electronic mail messages triggered by the arrival of new content into the MEDLINE index as the method of obtaining the enriched bibliographic data.

7.3. Delivery of Print Products

The quality of the scanned TIFF images provided by AT&T varies greatly, although the image cleanup algorithms help substantially. In any event, the logistics involved with transporting the higher resolution images across the network, even for local transfers, has the potential to create a bottleneck. One solution that we have been examining is to designate an attribute associated with a URL, perhaps in accordance with the evolving URI standards or the DIENST [Davis 94] bibliographic content retrieval protocol, that would allow an HTTP query to request a printable instance of the object referred to by the URL. Providing compressed PostScript data over the network would both lessen the bandwidth consumption as well as provide much higher quality printed output.

7.4. Full MeSH

The authorized, hierarchical Medical Subject Headings provided by the NLM is a standardized vocabulary for conducting uniform biomedical bibliographic searches on the MEDLINE index. We are currently in negotiation with the DLA procure a machine readable copy of this database so that we can integrate it into our searching and alerting systems. Such a controlled vocabulary would provide consistency between the Red Sage content and the MELVYL MEDLINE system.

7.5. LINK Tags

The specifications for HTML+ include the facility to impose an order on a set of HTML pages using the <LINK> and <GROUP> tags. At the CKM, we are implementing this functionality, as it lends itself especially well to this application. We intend to create relationships between pages in an article that fit into the defined roles of NEXT and PREVIOUS, as well as to the article's meta-information, such as CONTENTS and PARENT. This functionality is being implemented directly into the Mosaic browser, so that the CGI scripts need only specify <LINK> tags instead of using hypertext links. We hope to use this as a method of printing whole articles, as the meta-information for each page relative to the article will be available to the server using the PARTOF value for the ROLE attribute of the tag.

REFERENCES

[Davis 94] Davis, J.R., Lagoze, C. Dienst, a Protocol for a Distributed Digital Document Library

[Kahle 89] Kahle, B. Wide Area Information Server Concepts. Thinking Machines, Inc. 3 Nov 1989.

[Lucier 90] Lucier, R.E. Knowledge management: refining roles in scientific communication. EDUCOM Review, Fall 1990, vol 25, (3):21-7.

[Lucier 92] Lucier, R.E. Towards a knowledge management environment: a strategic framework. EDUCOM Review, Nov.-Dec. 1992, vol 27 (6):24-31.

[McCool 94] NCSA httpd Overview.

[McCool 94a] The Common Gateway Interface

[O'Gorman 92] Image and document processing techniques for the RightPages electronic library system/IN:Proceedings, 11th IAPR International Conference on Pattern Recognition, Vol II. Conference B: Pattern Recognition Methodology and Systems. The Hague, Netherlands, 30 Aug-2 Sep, 1992, Los Alamitos, CA, USA; IEEE Computer Soc Press, 1992, p260-3.

[Story 92] Story, G.A., O'Gorman, L., Fox, D., Schaper, L.L., et al, The RightPages image-based electronic library for altering and browsing, Computer, Sept. 1992, vol 25, (9); 17-26.

[Wall 90] Wall, L., Schwartz, R.L., Programming Perl, O'Reilly and Associates, Inc, 1990.

[email protected]

Marc Salomon is a Software Engineer with the Innovative Software and Systems Group of the Library and Center for Knowledge Management at the University of California, San Francisco Medical Center. Most recently, he has been working on both the Red Sage Electronic Journal Project as outlined in this paper, and with the Galen II project that will use the World Wide Web as the base technology for the next generation of library and informatics presentation and delivery.

EDUCATION

B.A. The University of Texas at Austin, 1989, Political Science/Latin American Studies, CS.

EXPERIENCE

1993-present, Software Engineer, ISSG/LIB-CKM, UCSF.

1991-1992, Senior Software Engineer, Carlyle Systems (now Marcorp), San Mateo, CA. Designed, wrote, implemented and debugged Online Public Access Catalog in C/INGRES on SunOS 4.1.x .

1989-1992, Software Engineer, Wang Laboratories, Redwood City, CA. Responsible for development of SCO UNIX-based word processing package.

1986-1989, Programmer Analyst, Department of Chemistry, The University of Texas, Austin, Austin, TX. Developed DOS/C mass spectrometer and temperature ramp control and data acquisition package.

1984-1986, Computer Operator, Clark, Thomas, Winters and Newton Attorneys at Law, Austin, TX. Supervised batch processing and developed report generation scripts.

1981-1984 Computer Operator, Mobil Exploration and Producing Services, Inc. Operated CDC Mainframes, wrote operator utilities, trained senior career operators on internal operating system structure.

1976-1980 Programmer, On-Trak Data Systems, Inc. Series of contracts while in high school. Wrote file filters in BASIC for IBM 5110 pre-microcomputers. AWARDS 1st place, 1977 Southern Methodist University programming contest. 3rd place, 1978 Southern Methodist University programming contest. 2nd place, 1979 North Texas State University programming contest.

David C. Martin is Assistant Director of the UCSF Center for Knowledge Management where he leads the software development team. The Center is building a knowledge management environment for the campus community. Prior to coming to UCSF, Mr. Martin worked for Molecular Simulations, Sun Microsystems and other silicon valley companies in the areas of user-interfaces, networking, artificial intelligence, databases and operating systems. Mr. Martin received his M.S. in computer science from the University of Wisconsin - Madison and his B.A. in sociology and computer science from the University of California - Berkeley.