College and Research Libraries ROBIN W. MACDONALD and J. MCREE ELROD An Approach to Developing Computer Catalogs A method of developing computer catalogs is proposed which does not require unit card conversion but rather the accumulation of data from operating programs. It is proposed that the bibliographic and finding functions of the catalog be separated, with the latter being the first automated. Such automation is seen as being advantageous on a cost basis. cARD CATALOGS WILL INEVITABLY BE RE- PLACED with a computer system. If the costs of card catalogs continue to rise and eventually meet the declining costs of computer usage, then at some point in this decade, economic pressures and other factors will force libraries to. abandon card systems for computer sys- tems.! It is not difficult to support . this hy- pothesis. One need only examine the operating costs of almost any library sys- tem over the past few years. The costs of creating and maintaining card cata- logs represent a significant portion of most academic library budgets; as cata- logs increase in size along with collec- tions and multiply with decentraliza- tion, these costs increase out of propor- tion to the growth of the library's total budget. In dealing with this phenome- non, libraries have introduced many im- provements in catalog methodology, but few are yet preparing for a major con- version to a computer-based catalog. Robin W. MacDonald is coordinator of technical services and ] . McRee Elrod is head of the cataloging divisions, University of British Columbia Library, Vancouver, B.C. 202 I One major reason many libraries are hesitant to prepare for this anticipated change is the cost of converting existing catalogs to machine-readable form. Al- though the cost of conversion is signifi- cant, it seems less when compared with the eventual costs of storing informa- tion, of computer time to service the in- quiries, of the many remote inquiry ter- minals needed for access and the other related costs of developing and main- taining the necessary software systems. Emphasis on the cost of conversion have tended to obscure other more important issues. Any consideration of the eventual re- placement of card catalogs should begin with the purposes of a catalog. A basic purpose is to put the user in touch with information, either by providing hold- ings information or by providing a de- tailed bibliographic description of spe- cific works. The first is a finding func- tion, to find what works are available, or to find a known work. The second is a bibliographic function, to provide an exact description of items held by the library. 2 Traditionally these two functions are combined in card catalogs and they re- side together fairly well. The relative 1 J 1 j importance of these two functions, the difference between them, and the possi- bility of separating them, are significant when considering an approach to devel- oping computer catalogs. A full biblio- graphic description is needed for one, yet a brief author and title will usually suffice for the other. Studies have indi- cated that many users, and most catalog queries, need finding information about possible or specific works. 3 With card catalogs, the cost of pro- viding one function along with the oth- er has not represented a significant ad- ditional cost; once entered into the in- formation store, the continuing cost to the library remains more or less constant regardless of the amount of informa- tion entered. With computer systems the amount of information stored represents a cost variable: the quantity of information entered will not only affect storage costs on a continuing basis but will also indi- rectly affect the costs of accessing the catalog. For each inquiry the amount of information processed will affect the time for accessing a given entry, the time to read, transmit and display the information for the user, and will also affect the requirements of the remote communication device used. All these factors contribute to the cost of using the system. The cost of using card cata- logs, on the other hand, is generally not a concern of the library, but rather a cost the user must bear. With computer catalogs, the cost of use will be boJne by the library budget. Another reason for examining the costs of card catalog systems in contrast to computer systems is the cost of de- centralization, where collections are di- vided among a variety of branch li- braries, resulting in increased costs. Studies of catalog costs at the U niversi- ty of British Columbia have shown that nearly 40 percent of the cards produced, sorted, and filed into the public cata- logs are required for branch library Developing Computer Catalogs I 203 catalogs, and since a union catalog is maintained these are all duplicate card sets. With computer systems, much of this redundant catalog data found in duplicate catalog files can be eliminat- ed by centralized data store with de- centralized access, and this access need not be restricted to library locations. The increasing difficulty of using large card catalogs augments the need for computer systems with their greater flexibility. Card catalogs, generally diffi- cult and costly to revise, are particularly inflexible for revising the structure of subject headings. Subject access is im- portant to many users, but some parts of large card catalogs become nearly un- usable when as many as a thousand en- tries fall under one heading. It becomes expensive and. difficult to introduce changes locally or to follow revisions made by the Library of Congress in the subject-heading structure. Some relief for this can be provided by divided catalogs with subject guide cards; but each time a change or rearrangement is introduced, the complexity of manual filing may increase, and the user is still left with an inflexible situation as well as a subarrangement of entries that doesn't help find what is wanted. Search options would be more numerous in a computerized catalog, and revision easi- er. Most importantly, they will be ulti- mately less costly than people. These pros and cons have been dis- cussed before, but always with the com- plete unit-card-by-unit-card conversion in mind. Are there no alternatives to perpetuating the card catalog on mag- netic tape? Perhaps, as a first step, it was necessary to explore that possibility. That exploration produced much val- uable information, stimulated much thought, and established an important standard in the MARC format. Do li- braries actually need a complete bib- liographic tool at the immediate dis- posal of every patron? Or do libraries need better finding devices to satisfy the 204 I College & Research Libraries • May 1973 most frequent demands made of the catalog? We believe there is a way to obtain a better finding device as a pre- liminary step in catalog conversion, without compromising the traditional bibliographic standard and without ex- cluding the possibility of providing a computerized bibliographic tool at some point in the future. If the two principal functions of the card catalog are separated, and dealt with as separate but interrelated re- quirements, the finding function could be served by a computer-based system, while the bibliographic function could continue to be served by a card file. Taking this approach, it is possible to begin by providing a computer inven- tory record: a file that would be linked to the circulation system, and would an- swer queries regarding what the library has, where it is located, and if it is avail- able. There would be only one point of access, probably by call number, which users would have to find by using the existing card catalog. The information required for this initial file would be similar to the information now used for most automated circulation systems. The file could in many cases be estab- lished from circulation records where these are in machine-readable form. Ad- ditional sources of information for such a machine file might be automated acquisitions and serials systems. The systems design for beginning an on-line cataloging system would be fair- ly critical; it would be important that the software, access methods, and basic design of the system allow the begin- ning file to become the central file to an eventual catalog system. Specifically, it would be necessary to be able to "build" additional files in connection with the central file, to provide additional ap- proaches. Systems already do exist that provide multiple access points using sep- arate "index" files linked to a "master" file. It is also important that the system be on-line because the introduction of this mass of additional data would out- strip the present batch-processing print- out methods. Once the central file is established, the next step would be to determine the or- der in which the various access points other than call number should be de- veloped. The choice would depend upon the particular library, and could begin with an area not already covered by the existing card catalog, such as an index to the classification system used. This, perhaps, is not the easiest alterna- tive, but it is attractive in that the cen- tral file would likely be in call-number sequence, and an additional subject ac- cess device not commonly found in North American libraries would be pro- vided. Alternately, the standard access points of author, title, or subject could be considered. The conversion method in each case would depend somewhat on the files available for the conversion process, but essentially the various au- thority files would be converted. Each "heading" would be entered once only and the related call numbers entered and linked to the central file. Where di- vided catalogs are used, it would be pos- sible to begin converting at the start of a particular file, and gradually replace the card-catalog access with a computer system. For many academic libraries, subject access would probably be the first choice in terms of the number of users that would benefit. A subject file might be established by using the LC subject headings on magnetic tape as a begin- ning source file, the seventh edition being currently available in a MARC format, with the prospect of the supple- ments also soon becoming available. Re- placement of that part of the card cata- log immediately necessitates considera- tion of those users who want biblio- graphic detail; it would then be neces- sary to provide bibliographic informa- tion in some form. One approach would be to provide a card shelf list. This, of course, would compel users who need a bibliographic description to use one system to find the item, and a second system to obtain fur- ther bibliographic ' detail. At the same time there would have to be an improve- ment in services to users of the finding process. This improvement would be- come a faster, easier-to-use, locating tool, accessible from many locations, easily and quickly revised and updated. 4 It would eliminate some redundancy in existing card systems and provide addi- tional benefits to users and the library not possible with card systems. This approach obviously requires se- lection of information to appear in the automated system, and of provisions made for those alphabets, symbols, etc., that cannot be reproduced on the stan- dard computer auxiliary equipment nor the library network of inquiry devices. These are not easy decisions; much is unknown about catalog usage and which elements of information are most use- ful. Those items now romanized on LC cards for nonroman alphabet materials might be used in the finding device: call number, author, title, and imprint date. These elements are the minimum re- quired for filing in the traditional card catalog-hence their romanization. Not only must user needs be consid- ered, but also the needs of the library's processing staff. In order to identify an item as an added copy or volume, the minimum bibliographic data usually re- quired is author, title, imprint, colla- tion, and series. This amount of biblio- graphic detail might not be included in the machine system; thus, the addition of a copy or volume might require the consultation of the classed card file by a library assistant, but no notation would need to be made there. In a large system, this would mean that a copy of the classed file might be needed in both public and work areas. Occasionally adding a slightly different edition as a copy by guess, it would be possible to Developing Computer Catalogs I 205 add copies without reference to the card file. This could be done by including the number of pages in the data entered on the 1nachine system. Rarely would two different editions be published in the same year, with the same number of pages (other than the simultaneous two- country publication of the same title, which is already ignored for added copy purposes in many libraries ) . The card copy would, in any event, match copy one. The number of pages is also rele- vant on the machine file for thickness of books when dealing with stack man- agement on an automated basis. The approach outlined here would change many processing operations, both during and after development. Subprofessionals, who formerly com- pared Library of Congress cards with books, might be comparing MARC in- formation on a display device with a book in hand and selecting data for the machine catalog; professionals, who formerly prepared worksheets for typists, instead might be entering infor- mation directly on a machine system. A shift in the skills required in processing would occur, along with a possible shift in the professional/ subprofessional ra- tio. Fewer, more highly skilled subpro- fessionals might be required once the library is no longer producing and filing cards in such large quantities. 5 The minority of users requiring full bibliographic information in such a system would still be consulting a card file (probably by call number). Not only would this file be the only source of complete bibliographic information, but it would also be the only place in which nonroman scripts would be shown in their original form. Oriental, Greek, Slavic, and Sanskrit materials would appear in the machine system in romanization only. This raises the issue of the signifi- cance of the MARC developments rela- tive to this approach. Although individ- ual libraries should compromise on the ·200 I College & Research Libraries • May 1973 availability of a standard bibliographic description, MARC developments should not necessarily make this kind of com- promise as well. Rather, a full conver- sion should be made at one location and local library systems would expect to ex- tract data needed for the local system. The availability of the full bibliographic data should continue to be a national concern, and the frequently-proposed National Network would conceivably maintain this information in a central store, or in regional centers. Those users requiring bibliographic details could connect through the local library system to access a data file locat- ed at some distant center. The manner of storing the full bibliographic data might include some form of graphic representation, rather than digital, for some languages, characters or symbols that may never be compatible with available, general-purpose computing devices. 6 Although data stored graphically could not be manipulated as could bib- liographic data in digital form, it could be called up for display by a machine system. Thus the bibliographic func- tion needed by some users, a service weakened earlier, could be recovered and probably improved. The costs of conversion, storage, proc- essing, etc., are not individually less on this basis, but collectively, the total cost would be a great deal less than if each and every library maintained the full data. Without pursuing all the implica- tions or requirements of a National Network on this basis, considerable study of user needs should be under- taken before massive retrospective con- versions are made. 7 Also, a realistic eval- uation should be made of the technical and economic limitations for computer reproduction of the many characters and symbols needed to represent Orien- tal, Slavic, Sanskrit, and other nonro- man alphabets. Even now, libraries are unable to influence the manufacture of a good circulation terminal. The un- pleasant truth must be faced: libraries simply don't have the influence to en- sure that the necessary tools will be available when needed. If specialized devices are essential, then their acquisi- tion should be examined before under- taking any long-range plans. Also need- ed are large budgetary increases over a period of time to cover the cost of com- puter time, terminals, networks, and other costs. This information system would de- velop from a manual system through a mixed system to an automated system. During its development the system might be thought of as existing for a time on each of three plateaus. The first plateau continues parts of the exist- ing catalog-author I title and subject files. The circulation system, serial rec- ord, and shelf list have merged into one on-line machine system, accessed by call number. Patrons would determine the call number from the card catalog, then use it to query a keyboard terminal to determine location and availability. The system responds, possibly with a visual display, giving volume, copy, location, and status information. The system ef- ficiently conveys the addition of vol- umes and copies, the placing of titles on reserve, the loss or loan of items, or the removal of material for binding. At the second plateau the traditional card catalog has been replaced. The call · number has served to link information entered from a variety of sources. From the authority file or author I title file all names (personal, corporate, and series ) have been entered along with the call number associated with them. From the subject catalog all subject entries have been entered along with their call num- bers. Titles have entered the system through a combination of circulation and acquisition history tapes. (Earlier an index to the classification might have provided a different type of subject con- trol. Terminals, consisting of typewrit- I I \ l er-like keyboards with visual display screens, may be queried by entering au- thor, title, series, or subject. Call num- bers with brief identifying data and holding and status information as de- tailed in the description of the first plateau would be displayed. Those desiring complete bibliograph- ic data would consult a card file by call number. Here is filed one card for each bibliographic item in the collection. For roman alphabet materials this might be a computer-produced card based on MARC tape; for nonroman script ma- terials, a Library of ·congress card; and locally produced cards for originally cataloged material. The method for es- tablishing this file might be by cards electrostatically printed from micro- film of the former shelf list. Copy and location information would no longer be entered in the public or work-area card files. For large or decentralized library sys- tems, the bibliographic file might be kept centrally in card form and avail- able at many locations on a microform, probably cartridge microfilm. At least one copy of this file would be kept in hard-copy form, the location and ar- · rangement depending upon the particu- lar library's requirements. The final plateau would replace this file by direct access to a central data bank. The existence of material unique to the local collection, or reluctance to see the last vestige of the three by five inch card catalog vanish, might allow at least one of the classed card files _ to be maintained. The filing of but one card per bibliographic item, and that by the rapid mode of call number, would aid the processing staff as would the ease for making changes in the system.8 The final product, then, would func- tion much like the computerized catalog as usually conceived. It would not, how- ever, be achieved by. unit card conver- sion. Nor would the data be stored in bibliographic units, but rather in a va- Developing Computer Catalogs I 201 riety of files (such as name authority, subject authority, and title) linked by call numbers. The local library would escape the high cost of storing and ma- nipulating the great mass of data re- quired to translate present card catalogs into machine form. Some of the current concern about catalog-filing problems could be elimi- nated by developing various access files, so that entries are inserted using human judgment rather than programming log- ic. The basic file initially would be cre- ated from preordered entries, i.e., the order of the particular file being con- verted. Then as new entries are added to these authority files, they would be inserted by catalogers in their proper place, perhaps by instructing the system to file the new entry between two others. At any rate, it is unlikely that program- ming of complex filing rules would be essential; however, it may be desirable to have the system "place" new entries initially and have the placement con- firmed by a filer or cataloger, with the ability to shift the logical position · of an entry. Since accessing the file would be controlled by a software system, the logical or physical placement of entries may not be as rigid or as important as it is with card systems; on the other hand, there may not be a need to depart from the existing structure of catalogs. An understandable, logical arrangement should be the main objective. This proposal requires study and de- velopment, particularly in regard to the needs of libraries for National Union Catalogs, regional and national net- works, shared cataloging, and MARC developments. The idea should be pur- sued, however, as an alternative to the seeming impossibility of each library ever having a full MARC catalog file. REFERENCES 1. Dolby states that "unless t.he storage, trans- mission, and retrieval of information in li- brary archives is automated, there is no hope 208 I College & Research Libraries • May 1973 of keeping pace with the exponential growth of libraries." J. L. Dolby, et al., The Cost of Maintaining and Updating Library Card Catalogs: Final Report (U.S. Office of Ed- ucation, Bureau of Resources [Washing- ton, D.C.: 1969]). 2. These two functions have been frequently distinguished, but it has usually been as- sumed that they must be combined in a computerized catalog. Cf. Robert Hayes and Joseph Becker, Handbook of Data Process- ing for Libraries ( New York: Becker & Hayes, 1970). On p. 587 the authors assert: "A catalog is usually regarded as being a very different thing from an index ... cata- logs providing very full and ·complete data and indexes providing very brief data." The idea of automating the index function of the catalog seems not to have occurred to the authors. 3. Catalog-use studies were made at the Uni- versity of British Columbia in preparation for the division of the dictionary catalog into author/title and subject files, and again in preparation for the imprint date filing of subject entries behind their guide cards. 4. Peter Simmons has shown the advantage of such ease of updating in the reserve book procedure, suggesting "faster responses to changing conditions with less human effort" in the adjusting of loan periods for books un- der heavy demand. "Reserve Collections," Canadian Library Journal 29:86 ( Mar.-Apr. 1972). 5. The average card set at UBC is now four- teen. Each year, $35,000 is spent maintain- ing the library's official author/title file alone. 6. Conference on Microform Utilization, the Academic Library Environment, Denver, 1970, Microform Utilization: The Academ- ic Library Environment (Denver, Colo.: University of Denver, 1971) p. 55. 7. Ben Ami Lipetz, User Requirements in Iden- tifying Desired W arks in a Large Library (New Haven: Yale University Library, 1970 ). 8. Time studies at UBC have shown that al- phabetic sorting and filing is more time-con- suming and therefore more expensive than call-number sorting and filing. \ ) J 1 j ~'