RICHARD DE GENNARO Harvard University's Widener Library Shelflist Conversion and Publication Program This paper briefly reviews the essential background and reports on the progress, evolution and future outlook of this unique program. It then describes in some detail the present and potential uses that can be made of the shelflist of a major research library after it has been con­ verted to machine readable form and can be manipulated by com­ puter. The uses fall in.to two broad categories: publications and read­ er services, and library management and statistical uses. It is a non­ technical article for librarians and library users. INTRODUCTION shelflist project were largely concerned with the strategy and the techniques of · IN 1964 WIDENER LmRARY, the central converting this large and unique file into research collection of Harvard Univer­ machine readable form. This paper will sity, developed a system for converting stress the present and potential uses that its manuscript sheaf shelflists to machine can be made of the shelflist of a major readable form and embarked on a proj­ research library after that conversion (or ect to computerize the 1.6, million entries a large part of it) has been completed. in the list. To date, more than 600,000 Two main categories of uses will be dis­ records from some of the most active cussed: ( 1) the production of publica­ classes in the library have been con­ tions of various kinds and the provision verted and used in various ways, and the of other reader services; and ( 2) library project continues as an accepted and im­ management uses including the genera­ portant part of the library's automation tion of statistical and other data for fur­ operations. The project has now come ther automation, for managerial pur­ of age; its feasibility and usefulness are poses, and for general research. firmly established and it seems appro­ priate at this time to review briefly the REVIEW OF THE PROJECT essential background of the program and The justification for embarking on the to report on its progress, evolution, and ambitious project to convert the estimat­ future outlook. ed 1.6 million handwritten entries in the The previous papers on the Widener old loose-leaf sheaf shelflists in Widener Library can be found in an article en­ At the time this paper was written, Dr. titled "A Computer Produced Shelflist" De Gennaro was Senior Associate University which appeared in C RL in 1965.1 The Librarian, Harvard University. He is now project was placed in the larger context Director of Libraries, University of Penn­ of the Harvard Library's overall automa­ sylvania. tion program in another article, "Auto- 318 / Con version and Publication Program I 319 mation in the Harvard College Library" a full-scale conversion and publication 2 .... , which was published in 1968. A tech­ program was begun in 1965. nical description of the operation in its The initial system was somewhat early stages was written by Foster M. primitive, with input and output limited Palmer in 1966.3 No detailed technical to the standard uppercase character set descriptions of the computer systems that was then commonly available on have been published since that time, al­ computer print chains. In June 1966 the though specific information can be ob­ system was improved so that the input tained from internal working documents. could be coded with an expanded char­ The preparation for publication of tech­ acter set to produce output with both nical descriptions of a rapidly evolving upper- and lowercase letters and the re­ system of a local nature is time-consum­ quired diacritical marks. The output for ing and difficult to justify. This article the published volumes continued to be will merely sketch in enough of the produced by photo offset from a com­ project's background to make it compre­ puter printout until further improve­ hensible without reference to the earlier ments in the system made it possible, papers. No technical material will be in­ late in 1969, to produce ·graphic arts cluded. quality printer's copy in double columns A library shelflist is a record of the by computerized photocomposition tech­ books arranged in the order in which niques. The evolution of the output for­ they appear on the shelves. It is main­ mat is virtually complete; all further im­ tained primarily as a tool for assigning provements will be in the input, process­ new and unique numbers to books that ing systems, and development of new are added to the collection and as an in­ products. ventory record of the books in a library. To date, more than 600,000 entries of Since the book collections in most Amer­ the total 1.6 million in the shelflist have ican libraries are arranged in classified been converted. Twenty-two volumes order, the shelflist is potentially useful to have been published in the Widener Li­ scholars, particularly if it can be made brary Shelflist series and a dozen more available in convenient form and if clas­ are scheduled for publication in 1970. sification schedules and author and title An estimated twenty-five to thirty ad- . indexes are provided. For most libraries ditional volumes will be required to com­ the maintenance of a shelflist is a routine plete the series. Several of the classes process and involves merely filing a copy that were initially keypunched in the of each main-entry card into the card limited uppercase format have been con­ shelflist in call-number order. However, verted by a combination of computer the Widener shelflist, for historical rea­ program and manual editing to the new sons, is largely handwritten in loose-leaf standard expanded character set and volumes, rather than on cards, and is format, while the rest will be completed therefore difficult to use and maintain. by the end of the year. Thus, all the In 1964 it became evident that, through records in the system will soon be in a the use of computer technology, the li­ single uniform and compatible machine brary could modernize the shelflist main­ format. The master files are arranged tenance procedure and at the same time in classified or shelflist sequence on make an expanded version of the shelf­ magnetic tape. Widener call numbers list available as an additional approach are machine processable and, since the to the library's holdings. Accordingly, an numbers are unique, they also serve as experimental system was designed to con­ identification numbers for the machine vert the shelflist to machine readable records. form and, after a successful pilot project, The entries in the old manuscript 320 I College & Research Libraries • September 1970 shelflists are not bibliographically com­ by librarians trained as systems analysts. plete. They were limited to call num­ The routine computer work has been ber, volume count, author, title, place, done for the most part on an IBM 1401 and date of publication. Frequently the which has 8,000 positions of core storage author's forenames were not spelled out and four tape drives and is located in the and the titles were shortened. Notes, library. In 1970 the 1401 will be phased I added entries, and subject headings were out after the entire system has been re­ not included. The strategy of the con­ designed and reprogrammed to run on """" version project is to accept the entry es­ an IBM 360-65 located in the U niver­ sentially as it is with some few excep­ sity's Computing Center. The system tions; obvious errors are corrected, au­ conversion will be done by the library's thors' full names are added when easily data processing staff. The occasion will obtainable, abbreviations in titles are be seized to convert the local shelflist spelled out, and a language code and a system into a more permanent and stan­ code distinguishing serials from mono­ dardized system based on the MARC graphs are added. All elements present II format. When the present system are tagged so as to permit machine ma­ was designed, the MARC II standard nipulation. The average number of char­ format for bibliographic entries in ma­ acters per record is 100, while full LC chine readable form did not exist. That records are estimated at 350-450 char­ format has now been completed and acters. This enforced limitation on the widely accepted internationally, and quality and completeness of the records programs are being written at several is unfortunate for many reasons, but it centers to manipulate bibliographic data has made the conversion and publication in that format in various wavs and for projects economically and technically various purposes. feasible. Had the shelflist contained Although Harvard shelflist entries are complete bibliographical records, the not as complete as full LC MARC II en­ project would not have been attempted, tries, the elements that are present can for various reasons. be tagged and put into the format, and Since clean and accurate copies of the those that are not can be left blank. Widener classification schedules are a When the library develops a system to necessary prerequisite for the prepara­ input its current cataloging in the MARC tion of the published shelflists, a major II fonnat, those entries can be integrated program was undertaken in 1966 to re­ into the new shelflist system, since the vise and edit the schedules. The sched­ machine format of the two kinds of en­ ules are being converted to machine tries will be compatible even though readable form, and a computer program they differ in the amount of data in­ nsed to facilitate editing as well as to cluded. format them into the two distinct forms In the more distant future it is ex­ that are required for the published lists.