Related Efforts in the Community
The XOBIS approach differs in many ways from that found in current MARC structures and the Anglo-American cataloging rules.It also differs from other similar, but less comprehensive approaches to marking-up library information.We attempt to monitor developments in these other approaches and track them on the Medlane Project website (6).Cursory remarks about a few of these related efforts provide some context for XOBIS.
The National Library of Medicine (NLM) has demonstrated international leadership in its early, extensive, and rapid adoption of XML both as a format for dissemination of its millions of records and as an internal communications format (7-10). This began in 1999 with the creation of MEDLINE Document Type Definitions (DTDs), and continued more recently, with the development of a DTD for Medical Subject Headings (MeSH). Notably, in the transition to XML, the MEDLINE DTDs were enhanced to better support inter-record relationships. Although these are stunning developments, NLM's effort has focused on a unique library's special needs with a focus on indexing records, which are less complex than MARC records, and the result is thus not readily adoptable by other libraries.
The Library of Congress (LC) has a long history of work with mark-up languages (11).Its recent flurry of XML activity further recognizes XML's growing importance in the library world.After exploration of SGML's possibilities from 1995-1998, LC's initial XML schema consisted of a literal mapping of each field in MARC to a counterpart element in XML, with each indicator becoming an attribute of that element. MODS, which appeared in 2002, regrouped some MARC fields and notably burst encoded fixed field values into meaningful terms (12). Similar to Dublin Core, ONIX from the publishing industry, and the Open Archives Initiative schema, MODS only covers a subset of MARC tags, lumps some elements, and is apparently intended for manipulating subsets of records for limited purposes (13-15). In June 2002, the official MARC XML format appeared, which is remarkably similar to the initial LC literal effort (16).Simply encasing MARC in XML does not take advantage of the strategic opportunity that XML affords libraries. It remains to be seen how useful this, or any of several other literal mappings of MARC, can be.
International interest has remained high since Lam's work in Hong Kong and the French BiblioML, released in 1999 (17-18). Korean acceptance has been phenomenal, especially since the selection of XML as the standard for electronic documentation for an e-Government project in 1999. A recent paper from Portugal is only the latest example indicating XML's value to libraries in a shrinking world (19).
MARC and cataloging rules have been under scrutiny recently, particularly in the Functional Requirements of Bibliographic Records (FRBR) (20-23).These admirable efforts provide an immense amount of information regarding the complex structure of MARC and cataloging rules.They identify many core concepts and issues, and while exhaustive and very informative, they seem to take a more traditional approach than might be warranted in the coming era of digital libraries. Attempts to resolve tensions between name/subject and subject/form are indicative of limitations in current structures (24-25).
We have not yet tried to correlate XOBIS with the FRBR. The treatment of series in appears problematic, but FRBR's emphasis on discrete entities and relationships is encouraging. The XOBIS Work element should not be confused with Work of FRBR (from Work/Expression/Manifestation/Item), although Item is likely the same.The XOBIS Version element can accommodate editions with nearly identical content, such as novels, by representing them as versions on a "single" record, while editions with different content, such as textbooks, could have separate records as their content is not substitutable.We hope to explore such issues more fully during our initial implementation.
The Dublin Core (DC) spearheaded by OCLC is useful in emphasizing that documents should contain basic, discrete metadata (e.g. date, title), but in our view is not sufficiently detailed to accommodate bibliographic information adequately (13). Perhaps its most interesting lesson comes from adjusting to change. After millions of records were created, optional attributes were added to compensate for insufficient detail. More recently the recognition of overlap in its Creator, Contributor, and Publisher elements underscored the difficulties in altering a schema.
We are encouraged by many domain-specific efforts, such as HEAL, although they tend to be enumerative and prescriptive (26). Notable in this category is the Visual Resources Association's schema, which combines simplicity and the crisp delineation of fundamental elements, as well as better provision for relationships than is usual (27).It is also telling that many such efforts are not occurring in libraries. After a period of exploration and implementational analysis, we anticipate that emphasis will shift to coordination of the many related, exploratory efforts.