TULIP started in early 1991. University systems and library leaders at a number of schools had been talking with Elsevier to find a way to accelerate the development of large scale systems for the distribution in electronic form of traditional journal information - information presently found only in print. Elsevier Science was looking at the same question from the publisher’s side and was looking for experience on which to make strategic developmental and investment decisions, whether in search software, document delivery systems, PostScript or SGML database files or network development. During two Coalition for Networked Information (CNI) meetings in the spring of 1991, it was agreed that if ten or fifteen universities would commit to the same basic experiment, then a publisher could justify investing in the creation of a major test bed. University participants outlined a project and organized a group of universities on the spot and so TULIP was started. The universities were invited to submit project proposals, and Elsevier Science started to establish the technical and organizational framework necessary for such a large project. Ultimately, the TULIP program became operational in January 1993, and nine universities had decided to participate: Carnegie Mellon University, Cornell University, Georgia Institute of Technology, Massachusetts Institute of Technology, University of California (all campuses), University of Michigan, University of Tennessee, University of Washington and Virginia Polytechnic Institute and State University. The participating universities have in common strengths in the physical and engineering sciences. In looking within these disciplines for a target area, a field was preferred in which the researchers were comfortable with computer applications and had a higher than average installed base of workstations. An obvious choice might have been computer science itself, but these users were expected to be atypical in their computer facility, as to make it hard to generalize results to other disciplines. Material science provided a field in which there was both a sufficiently large corpus of frequently-cited material within one publishing company and interested faculties. TULIP files consist of the scanned page images plus bibliographic data and unedited, OCR generated, “raw” ASCII full text of 43 Elsevier and Pergamon materials science and engineering journals. The project started in January 1993 with the 1992 back file and continued for three additional years, totaling more than 500,000 pages.

TULIP is a cooperative research project testing systems for networked delivery and use of journals to the user desktop. The participants set three objectives at the outset:

Technical:

To determine the technical feasibility of networked distribution to and across institutions with varying levels of sophistication in their technical infrastructure. “Networked distribution” means sending the information both across the national Internet and over campus networks to the desktops of students and faculty. Elsevier will deliver the journal information to participating universities in standard formats. The universities will incorporate in local prototype or operational systems. A wide variety of delivery alternatives, search and retrieval systems and print-on-demand options will be compared.

Organizational and economic:

To understand through the implementation of prototypes, alternative costing, pricing, subscription and market models that may be “viable” in electronic distribution scenarios; comparing such models with existing print-then-distribute models; and understanding the role of campus organizational units under such scenarios. The overall goal is to reduce the unit cost of information delivery and retrieval. “Viable” means economically and functionally acceptable to all parties.

User behavior:

To study reader usage patterns under different distribution (technical, organizational and economic) situations. Improvement in the functionality of the information, whether as to article structure or retrieval tools, will also be considered. Certain data will be collected uniformly at all sites for analysis in the aggregate and for comparison among different systems.

The technical issues are discussed in chapter II. The organizational/economic issues are discussed in chapter V and user behavior is the subject of chapter IV.

The TULIP project, some highlights:

1991

• In March the TULIP project was initiated at the CNI Spring Task force Meeting.
• In the remainder of the year Elsevier Science made a choice of suppliers who could process the required material, and of the tools they would be using to do this.

1992

• Meeting between Elsevier Science management and university administrators to agree on the project’s goals and on the roles and contributions of the participants.
• Nine universities of the initial group of about 15 submitted a proposal to participate in the project.
• The open exchange format (=TULIP Technical Specifications) was specified and refined further, based on experiences and demands from universities.
• Purchase of equipment and development of production facilities at Elsevier Science took place to provide large-scale page scanning, OCR-ing and editing references.
• Purchase of equipment, setup of Internet facilities and development of Dataset loading, customization (=creating Dataset “subsets” per university) and subsequent delivery procedures took place at Engineering Information (Ei). This included development of an E-mail-driven facility for single article delivery.

1993

• First of the “regular” TULIP group meetings at the bi-annual CNI Task Force meetings, involving Elsevier Science staff, university participants and “observers” from the universities which had decided not to participate.
• Processing of TULIP data. Dispatch of material was split in two parallel streams:
• 1992 backlog material was directly dispatched on CD Rom to all universities.
• 1993 current material was dispatched by Ei over the Internet.
• This delivery of TULIP data was based on a so called “Push” mechanism in which Ei initiated the sending.
• At the universities development and/or adaptation of local library systems took place. The technological infrastructure at the universities determined to a large extent the technical solutions that were chosen: “hi-tech” X-Windows/UNIX based systems with which it was possible to view page images on screen, vs “low-end” Telnet based systems for less sophisticated workstations.
• The University of Michigan system went “live” in June.
• In the fall, the first technical meeting, uniting technical implementers from all the universities was held, hosted by the University of Michigan. This meeting proved very valuable in the exchange of ideas, which continued to take place on the TULIPTECH-L listserv.

1994

• At Elsevier Science, a stabilized production and distribution environment was established after the backlog was eliminated.
• Based on experiences, the quality and validation of the material (notably page images) was enhanced and a checksum facility was introduced.
• At Ei the initial “push” strategy of FTP deliveries was changed to a “kicked-off push” in which university staff initiated a Dataset delivery after ensuring that enough resources were available for receipt of a Dataset.
• A number of TULIP systems at the universities went “live”.
• In addition to numerous local promotional activities, Elsevier Science produced posters, banners, lapel pins, newsletters and a TULIP home page on the WWW to support the promotion of the project at the universities.
• Logfiles, the electronic records of user activity, started to de delivered to Elsevier Science.
• Universities started to investigate and/or implement MS-Windows and World Wide Web/Mosaic based systems, because the X-Windows platform was found to have too little “critical mass” in many places.
• Elsevier Science cranked up its internal initiatives to redesign its internal production operations and to develop an “electronic warehouse”.
• Elsevier Science started to scan all 1,100+ Elsevier journals based on the TULIP model for its Electronic Subscriptions project.
• The second technical meeting, hosted by Carnegie Mellon University, was held in the fall.

1995

• SGML bibliographic data files were added to TULIP Datasets, resulting in a moderate backlog at the scanning office.
• The number of TULIP journal titles increased from 43 to 83 journal titles, derived from the Electronic Subscriptions operation. Not all universities decided to take the additional titles.
• Because of problems with FTP deliveries resulting in large backlogs, it was decided to shift to CD Rom deliveries of Datasets.
• Focus groups and interviews, organized by Elsevier Science, took place at the University of Michigan, MIT and the University of California to research (changes in) user behavior and user needs as well as to have users evaluate the project. The University of Tennessee and Carnegie Mellon University also conducted focus group research.
• At the bi-annual TULIP group meeting, a separate discussion session on economic models was held.
• Most universities were busy investigating and developing World Wide Web functionality. A number of those developments were put into operation in 1995.
• At Elsevier Science, a new scanner was installed in the last quarter, resulting in significant improvements in the gray scale parts of the journals, such as photographs.
• Announcement of Elsevier Electronic Subscriptions, making all Elsevier Science journal information available in electronic form.
• The third and final technical meeting took place in the spring, hosted by Elsevier Science in Amsterdam. Besides the usual exchange of experiences and ideas, this meeting included an introduction to Elsevier Science publishing systems, and some brainstorming sessions on the next steps to be taken towards full scale implementations from a technical viewpoint.

1996

• Publication of the TULIP final report.
• TULIP is being succeeded by a full scale production commercial program, called Elsevier Electronic Subscriptions (EES). All of Elsevier’s 1,100 plus titles are available in electronic format. Elsevier Science is offering EES to a number of selected (“high end”) customers, among which obviously are the TULIP participants.
• Of the TULIP participants, some are going to a “full blown” EES implementation, meaning that they will receive all their subscriptions in both paper and electronic form. Others have found that they are not ready or willing to implement an EES based digital library on a full scale, due to technical, organizational or budget constraints. Finally, there are two instances in which the universities feel that they have not learnt enough, and have asked Elsevier for an extension of the experimental period.

Elsevier Science

With headquarters in Amsterdam, Elsevier Science has companies in Oxford, New York, Tokyo, Lausanne, Paris, Seoul, Rio de Janeiro and Shannon. Elsevier Science aims at linking scientists from all over the world through a comprehensive publishing program. Scientific progress is highly dependent on certified and refereed information and on the access to the information. Some of Elsevier’s hallmarks are close cooperation with authors and editors, support of the refereeing system, speedy preparation of material for publication, commitment to innovative products and services, expertise in international marketing and distribution and quality control throughout the process. Elsevier Science aims to advance science, technology, and medical science by fulfilling the communication needs specific to the international community of scientists and engineers, and associated professions. Elsevier Science wishes to collaborate in this with libraries anywhere in the world where possible.

Elsevier Science is committed to converting all of its production processes at its different publishing houses (over 1,100 journals and 1.8 million pages published in ten locations in a previously non-standardized way) to produce media-neutral databases based upon SGML tagging as its primary output, generating the print as well as electronic products from that electronic source-material. This fundamental change in Elsevier Science’s production, which involves changing almost all production procedures, hardware as well as software, and most importantly, hundreds of people’s jobs, is currently under way, to be completed at the end of 1996. In the meanwhile, Elsevier Science is providing all its journals in electronic form by scanning the paper journals, as was done in the TULIP project.

 Description of universities

The TULIP project unites nine of the leading universities in the USA. Although they have in common the drive to be at the forefront of the transition to digital libraries, they differ vastly in size, scope, resources, etc. On the one hand, very large institutions like the University of California system were participating in the TULIP project, on the other hand there were also much smaller institutions like the University of Tennessee participating. This diversity was reflected in the approaches of the projects, in their execution and in their results. Below is a short overview of the characteristics of each of the participants; more details can be found in the university reports (appendices I through IX).

Carnegie Mellon University

• Carnegie Mellon University has seven teaching colleges or schools, most of which offer undergraduate as well as graduate programs, as well as over 50 interdisciplinary research centers. The annual budget of the university is approx. $360 million, including funded research. The University Librarian is responsible for library operations at Carnegie Mellon and reports to the provost. The staff includes 36 professionals and 50 staff, plus a significant workforce of students, around 45 FTE in all. The University libraries’ collections include more than 852,000 volumes; over 900,000 microforms and graphics; and 3,850 periodical subscriptions housed in three locations. The Library Information Technology (LIT) department is responsible for day-to-day operations (eight FTE) as well as research and development (five FTE). The head of R&D is responsible to the head of LIT who reports to the University Librarian. In addition to TULIP, LIT is currently supporting several major technology R&D projects valued at around $2,000,000.

Cornell University

• Cornell University is partially public (it is the land grant university for the State of New York), and partially private. The university consists of thirteen colleges; eleven of these are in Ithaca, and two (the Medical College and the Graduate School of Medical Studies), are located in New York City. The Cornell University Library system contains 5.8 million volumes, 6.8 million microform items, as well as large collections of maps, videos, sound recordings, and computer files. The library system consists of nineteen libraries, two of which --the Engineering Library and the Physical Sciences Library-- were the primary Cornell participants in the TULIP project. The Engineering and Physical Sciences librarians report to an associate university librarian, who reports to the University Librarian. The University Librarian reports to the provost. Cornell Information Technologies (CIT) consists of approximately 325 staff members, divided into three divisions: (a) Information Resources, (b) Support Services and Academic Computing, and (c) Network Resources.

Georgia Institute of Technology (Georgia Tech)

• Georgia Institute of Technology has five colleges, plus the Georgia Tech Research Institute. Research is an integral part of the education process at Georgia Tech and has grown to over $168 million annually, including research in the colleges and at the Georgia Tech Research Institute. Research activities are diversified and are centered on areas where the nation has a vital interest - defense, manufacturing, health and the environment, and electronics. Students and faculty have access to a catalog collection of over 2,929,507 bibliographic units in the Price Gilbert Memorial Library. Technical reports number approximately 2,375,000. There are over 5,130,000 patents; government documents include over 623,000 publications and nearly 173,680 maps. Approximately seventy-five percent (75%) of the total collection is in scientific or technical fields. Literature searches and other reference services are provided from more than 500 bibliographic and factual data bases. There are about 100 library staff, of which eight are systems staff.

Massachusetts Institute of Technology (MIT)

• MIT is an independent, coeducational, privately endowed university, broadly organized into five academic schools, housing 21 academic departments, complemented by many interdepartmental laboratories, centers, and divisions. MIT has a total budget of approximately $1.4 billion and research revenues of approximately $700 million. Reporting to the provost and with a total staff of about 200, the MIT libraries’ resources comprise more than 21,000 current journals and periodicals and extensive back files totaling more than 2.3 million volumes. The library system is composed of five major (divisional) libraries as well as a number of smaller branch libraries.
• Reporting to the senior vice president and with a total staff of about 240, Information Systems (IS) provides comprehensive computing, networking, and telecommunications services for all of MIT. IS services and facilities support the academic, research, and administrative use of a broad scope of information and computing technology, and include responsibility for the Athena Computing Environment and MITnet, the campus-wide computer network. The Athena Computing Environment offers computing resources designed for educational use by students and faculty. MITnet connects thousands of computers across the campus and to the Internet.

The University of California

• The University of California is one of the largest institutions of higher education in the world. It has academic programs in over 150 disciplines, and ten percent of all Ph.D.’s awarded in the United States come from the University of California. Its library system contains over 100 libraries on the nine campuses, with collections totaling over 27 million volumes. The Division of Library Automation (DLA) is responsible for the University of California system-wide MELVYL system, a computer based library system providing access to a variety of bibliographic databases. DLA is also responsible for operating the inter-campus TCP/IP data communications network. In 1995 this network was upgraded from a network running at fractional T1 speeds to an SMDS network. One of the major justifications for this large increase in bandwidth was the need to adequately support applications like TULIP.

The University of Michigan

• The University of Michigan, located in Ann Arbor, contains 18 schools and colleges. During fiscal year 93-94, the University had grant research income of $386 million, the largest of any U.S. university. Although the university’s schools and colleges are managed through a decentralized organizational structure, there is a central agency for information technology, ITD (Information Technology Division), managed by the Vice-Provost for Information Technology. This unit provides basic IT infrastructure and support, maintains public computing sites, advises on hardware and software, and cooperates with others to meet University needs. Some of the schools and colleges have their own information technology support groups. This is the case for both the University Library which manages MIRLYN, the campus library management system, and the College of Engineering which oversees the Computer-Aided Engineering Network (CAEN). The School of Information and Library Studies (SILS) is also a contributor to the University’s development of electronic services. Consistent with Michigan’s tradition of interdisciplinary activities, ITD, the University Library, CAEN, and SILS are collaborating under the umbrella of the Digital Library program. TULIP, NSF-UMDL, a project funded by the National Science Foundation and others, and JSTOR are part of this University of Michigan Digital Library (UMDL) umbrella. The units are working together in development and production activities.

The University of Tennessee

• The University of Tennessee is the official land-grant institution for the State of Tennessee, with its main campus in Knoxville. The University offers more than 300 degree programs. Developments in graduate education have been accompanied by expanded cooperation with Oak Ridge National Laboratory (ORNL) and the Tennessee Valley Authority, and by growth of major research programs including those in the fields of energy, biotechnology, and robotics. The University of Tennessee, Knoxville libraries own approximately 2 million volumes, more than 3.5 million manuscripts, 2 million microforms, 32,000 audio and video recordings, plus United States and United Nations documents. The University of Tennessee libraries employ 185 full-time staff, and subscribe to more than 11,000 periodicals and other serial titles. TULIP was implemented at The University of Tennessee by a team of library staff that included an information sciences professor, independent of the computing and telecommunications services.

The University of Washington

• The University of Washington, a public research university with a main campus in Seattle and branch campuses in Tacoma and Bothell, has 16 major schools and colleges. The University of Washington libraries system is one of the largest research libraries in North America. Its collections exceed five million catalogued volumes, an equal number in microform, several million items in other formats, and more than 50,000 serial titles. The libraries system is composed of the Suzzallo and Allen Libraries with collections primarily in the humanities, natural sciences and social sciences, three other major libraries, 18 specialized branch libraries and two branch campus libraries. Computing at the University of Washington is the responsibility of Computing & Communications (C&C), which has experienced a more than 25 percent annual growth rate in active accounts in the last five years. There are now over 40,000 individuals using the the University of Washington Uniform Access computers. The services offered by C&C; encompass a wide array of computing and networking environments.

Virginia Polytechnic Institute and State University (Virginia Tech)

• Virginia Tech, the largest state supported institution of higher education in the Commonwealth of Virginia and a land-grant institution, has nine different colleges, the largest being the College of Engineering. The Virginia Tech libraries consist of a main library, three branch libraries at the main campus in Blacksburg, and another branch library in Fairfax, Va. There are six administrative units that comprise the University libraries - Administrative Services, Collection Development, Library Automation, Special Collections, Technical Services, and User Services. Library personnel consists of 40 professional librarians, 110 support staff, and approximately 63 FTE of student assistants, reporting to the Dean of University Libraries, who in turn reports to the Vice President for Information Systems. Within the University libraries all IT services, equipment installation and repairs, software and network maintenance, etc., are provided by the Library Automation Department. Campus-wide IT services are provided mainly by other information iystems groups, such as the Computing Center, Communications Network Services, and Media Services, but library personnel do participate in various IT outreach projects that affect a broad spectrum of the university community.

The project was implemented under different organizational structures at different universities. At all sites the libraries were involved, at some sites the computer centers; other technical organizations and library schools were involved in the project as well.

• At Carnegie Mellon University, the project started out as a joint activity between the Computing Services and the University libraries. This turned out to be an unworkable solution, because nobody was focussed on the effort and as a result the work did not get done. After about a year, TULIP became the full responsibility of the libraries, which allowed them to supply the necessary effort to the implementation of TULIP much more quickly than had been the case before.

• Cornell University had two separate groups involved in the project: the Cornell Information Technologies and the library. While the library provided administrative and public service support for TULIP, the main technical work on the project was done by a combination of special support from CIT and grant funding received to further the creation of the Cornell digital library. CIT also assumed responsibility for archiving the TULIP materials. Part way through the TULIP project, the library made the decision to divide organizationally those projects considered to be experimental, such as TULIP, from regular and routine online services (e.g. online catalog maintenance) already in production. This was done mainly in order to protect the support for basic online services. This decision had the advantage of clarifying reporting relationships, which resulted in significant improvements in productivity allowing Cornell to better guide the development of the TULIP effort.

• Georgia Institute of Technology managed the project with library staff exclusively, no computer center staff was involved. Georgia Tech has a fairly large systems staff incorporated within the library, dedicated to library projects, who really understand library issues. The implementation could be done “in-house”, so there were no organizational problems; as Georgia Tech already had staff support for the database on which the TULIP implementation was based, it was relatively easy to implement TULIP on this site.

• At the Massachusetts Institute of Technology, organization of the TULIP experiment took place within the context of the MIT Distributed Library Initiative (DLI), a collaborative effort led by the MIT Libraries and Information Systems. The TULIP implementation team was called BULB and was comprised of public services staff from the libraries, programmer analyst expertise from information systems, a research associate from the Dept. of Materials Science, and a counselor from the MIT Intellectual Property Office.

• At the University of California, the university libraries of all nine campuses were involved by means of an advisory/liaison group of librarians. The project was managed by the Division of Library Automation (DLA) at the office of the president, where all the technical implementation took place.

• As mentioned above, at the University of Michigan the project was carried out under the umbrella of a joint undertaking of the Information Technology Division, the College of Engineering, the School of Information and Library Studies and the University library. A joint team with representatives from each group guided developments through each step of the TULIP project.

• The University of Tennessee had involvement from the library only; a team of library staff with responsibility in public service, networking, and systems implemented the TULIP application. A professor from the School of Information Sciences was involved throughout the project and participated in promotion, training and research activities.

• The project at the University of Washington was rooted in the systems side of the libraries and the university information systems. The branch librarians dealing with the potential users were only marginally involved in the development of the application. The University of Washington abandoned their TULIP implementation two years into the project, because they wanted to move all development to the Web environment and because they felt the project did not have enough of a potential user base on their campus to make it a success.

• At Virginia Tech, the project was managed by the University libraries, while resources were provided by both the libraries and the Computer Center. User contact was maintained through an interested professor in the Department of Material Science and Engineering in the College of Engineering.

The total numbers of faculty and students per university are listed below and where available the primary target departments and their user groups for the TULIP project are also listed.

• Carnegie Mellon is a highly selective small, private, coeducational university with approximately 4,300 undergraduate students, 2,400 graduate students, and over 550 regular faculty. Some 30 faculty, 55 graduate students and 50 undergraduate students could be considered the primary target group for the TULIP material, while there is a group of about three times as many people who could have an interest in parts of the TULIP journals list.

• Cornell University has a current enrollment of 19,000 students (13,000 undergraduates and 6,000 graduate students), and a faculty of 1,600. The College of Engineering (218 faculty, 1,055 graduate students) as well as the science faculty and students in the College of Arts and Sciences, especially in the fields of physics (45 faculty, 190 graduate students, two research associates), and chemistry (31 faculty, 145 graduate students, 15 research associates) are the primary target groups for the TULIP content.

• Georgia Institute of Technology has about 5,000 faculty and staff, 4,000 graduate students and about 8,000 undergraduates. Of these, about 220 can be considered potential users of the TULIP material.

• Massachusetts Institute of Technology’s total enrollment is approximately 9,800 students, with somewhat fewer undergraduates than graduate students. The MIT faculty numbers approximately 1,100, with a total teaching staff of over 2,000. The potential user group of the TULIP files is estimated at about 450, the primary group being the Materials Research Department, but some of the material may be of interest for users in at least nine other departments.

• The University of California’s nine campuses comprise 167,000 students and 130,000 faculty and staff. They are composed of approximately 122,000 undergraduate and 45,000 graduate students, and 7,000 faculty and 32,000 other academic personnel. Of the nine campuses, eight have more or less substantial departments involved with material science/engineering. The total potential user group for the TULIP files at the University of California is almost 3,400, but many of these can not be considered potential “core users” of the TULIP files, as their primary research interests are not represented in the TULIP journal list.

• The University of Michigan has an enrollment of 37,000 students, of which 30% are graduate students. The College of Engineering contains the largest number of materials scientists as well as computer scientists and others interested in the TULIP titles. However, users can also be found in the physics and chemistry departments, and to a lesser extent in the Dental, Medical and Pharmacy schools. Altogether, a potential user community of 360 faculty and graduate students is estimated.

• The University of Tennessee has an enrollment of 26,000 students and employs 1,600 faculty. The TULIP audience included faculty and graduate students in the University of Tennessee, College of Engineering Department of Materials Science and Engineering, as well as ORNL researchers whose work involved materials science. There are 15 Materials Science faculty and approximately 100 graduate students. The Materials Science Department also has a small undergraduate program with 75-80 students enrolled.

• The University of Washington is populated by about 3,500 teaching and research faculty, about 13,000 staff and about 35,000 students, of which 8,000 are graduate students. The primary target group for the TULIP material would be in the College of Engineering, which offers education in traditional fields dealing with transportation systems, manufacturing and process industries in the Departments of Aeronautics and Astronautics, Civil Engineering, Chemical Engineering, Electrical Engineering and Mechanical Engineering, and also rapidly-growing new fields in bioengineering, computer science and engineering, materials science and engineering and technical communication.

• Virginia Tech has over 23,000 students enrolled in graduate and undergraduate education, while the College of Engineering alone has over 5,000 students enrolled. The primary target group for this project was to consist mainly of the faculty, graduate and undergraduate students in the area of materials science, consisting of the Materials Science and Engineering Department, with 20 faculty, approximately 50 graduate students and 50 undergraduate students, complemented by approximately 50 additional faculty and 50 graduate students from related fields such as Chemical Engineering, Chemistry, Engineering Science and Mechanics, Electrical Engineering, Mechanical Engineering, and Physics. The initial user group will be the MSE department.