Digital Libraries ‘95 Proceedings 147 June 11-13, 1995Digital Libraries: Issues and ArchitecturesPeter J. Nürnberg, Richard Furuta, John J. Leggett, Catherine C. Marshall, Frank M. Shipman IIICenter for the Study of Digital LibrariesTexas A&M UniversityCollege Station, TX 77843-3112 USA{pnuern, furuta, leggett, marshall, shipman}@bush.cs.tamu.eduABSTRACTThe research field of digital libraries must be viewed as aunion of subfields from a variety of domains combined withnew research issues in order to realize its full potential. Aclear exposition of the research issues involved has not yetbeen given. Most approaches to building digital librarysystems have thus far been limited to addressing specificdigital library problems as variations of problems from otherfields. This paper presents a taxonomy of digital libraryelements. Consideration of the elements in this taxonomyhelps suggest a variety of issues. Example elements andsome issues they suggest are used to populate the taxonomy.The paper continues by presenting a general digital librarysystem architecture. Issues suggested by the taxonomy areshown to have implications at many levels of digital librarysystem architectures for both design and implementation.This is illustrated by considering the implications of oneissue (personalizing presentations) at several architecturallevels and in the context of a set of current technologies.Keywords: digital library issues, digital library architecture,databases, physical libraries, World Wide WebINTRODUCTIONThe emerging field of digital libraries brings togetherparticipants from many existing areas of research. Currently,the field lacks a clear agenda independent of these otherareas. It is tempting for researchers to think that the field ofdigital libraries is a natural outgrowth of an already knownfield. From a database or information retrieval perspective,digital libraries may be seen as a form of federateddatabases. From a hypertext perspective the field of digitallibraries could seem like a particular application ofhypertext technology. From a wide-area information serviceperspective, digital libraries could appear to be one use ofthe World Wide Web. From a library science perspective,digital libraries might be seen as continuing a trend towardlibrary automation. There is some truth to these perspectives(as well as others) but none address the field as a whole andits research agenda. The field of digital libraries will belimited if viewed only as a subfield of prior researchinterests. To realize its full potential, the field must beviewed as a union of subfields from a variety of domainscombined with additional goals, and thus new researchissues. Digital library research must both respect theexisting tradition of our physical libraries and transcendcurrent practice in developing a new, broader researchagenda.What are the research issues central to digital libraries? Oneissue might be how to digitize objects and put them on-line.A second might be how to include new forms of informationthat do not have temporal or tangible representationnecessary for inclusion into physical libraries. Anothercould be how to locate materials in the new digital library.Yet another would be when to use and when to transcend theexisting technologies and traditions of the physical library inits digital form. Still other issues stem from the problems ofinformation overload created by new informationtechnologies. This paper presents a framework for thinkingabout the field of digital libraries and the research issues thatare part of it and demonstrates how these issues affectdigital library systems.The next section gives an analysis of the digital librariesfield by positing that the digital library can be modeled tosome degree after the physical library, and discussing therelationship between the two. In order to show the breadthof the research agenda in digital libraries, a taxonomy of theof elements of the digital library, and some issues raised byconsidering these elements is then presented. Followingthis, a general system architecture for digital library systemsis presented. Issues suggested by considering the priortaxonomy are shown to affect many layers of these systems.PHYSICAL AND DIGITAL LIBRARIESWhy is a digital library called a library at all? This questionhas been addressed by various members of this researchcommunity. Miksa and Doty [10] discussed the notions ofcollection, information sources, and place with respect tophysical libraries and how these notions might carry overinto the digital realm. Levy and Marshall [6] consideredhow work practices in physical libraries might be used in thedesign of digital libraries. The physical library can providethe starting point for discussing the elements and domains ofdigital libraries. An element of a library is a constituent partof the library. A domain of the library is the universe fromwhich the library materials are drawn.Digital Libraries ‘95 Proceedings 148 June 11-13, 1995Data Metadata ProcessesTranslations ofPhysical LibraryEntitiesBookJournalMovieStatic indexClassificationsSpatial arrangementAcquiring dataSuggesting sourcesHelping locate sourcesNew DigitalLibrary EntitiesHypernovelScientific visualizationComputer programDynamic indexPersonalized structureAnnotationsFull-text searchingPersonalizing presentationRetrieving by agentsFigure 1: Taxonomy of Digital Library Elements.ElementsIt is helpful to consider three broad classes of libraryelements: data, metadata, and processes. Data are librarymaterials. Metadata are information about the library and itsmaterials. Processes are active functions performed overlibrary elements. For example, a book in a library may bethought of as being data of that library. An index over booktitles (in a card catalog, for example) may be thought of aslibrary metadata. The act of a librarian helping a patron finda book by suggesting the use of the card catalog may bethought of as a process.This classification is vague, in the sense that it may bedifficult or impossible to classify any given library elementas distinctly belonging to a particular class. It may bepossible to view a single element as belonging to all threeclasses. However, this classification is useful since itprovides a framework for discussion about library elements.Physical library elements often fulfill some role for a givenlibrary user at a given moment. These roles often can beassigned in specific cases in a meaningful way.Because this