Defining metadata barely as "data about data", which is its literal meaning, is too simple considering the many dimensions this term offers. It is understood to mean structured data about resources, including resource description and discovery, as well as the management of information resources. Basically, a threefold division of metadata has been identified [Day98]:
This broad definition is due to the fact, that a wide range of communities make use of metadata, all with a subtly different approach. Libraries, archives, museums and other information and record-keeping communities, all use metadata for their catalogues, indices, and documentations. These again subdivide in an array of disciplines. Metadata has also become a fashionable term, and is often overused [HPD97]. As the various communities come together across boundaries clashes in understandings and definitions are often the case. Even though much activity is centred on development of standard formats for metadata, it cannot be viewed in isolation from the context in which it is used.
As two main approaches to metadata, (1) the library science oriented bibliographic control approach and (2) the computer science oriented data management approach can be discriminated [BNP97]. The former uses metadata primarily for resource description and discovery purposes, whereas the latter focuses on its administrative functionality for the management of data repositories.
As long as there are libraries or other collections metadata has been created in form of catalogues and indices. Increasingly, such metadata are being incorporated into digital information systems. For this reason, most popular, the MAchine Readable Cataloguing (MARC) formats encoding cataloguing rules have been defined. MARC refers to a family of formats (e.g. USMARC, UKMARC, UNIMARC) created for the exchange of bibliographic and other related information in machine readable form, containing a rich variety of elements related to resource identification and discovery. Originally developed by the Library of Congress, Washington DC, it has evolved to a de facto standard by now.
The computer science community primarily applies metadata to help administer and manage resources as well as for documentation. As an example, the highly detailed Australian Record-Keeping Metadata Schema (RKMS) was developed in this understanding of metadata. It is designed to reliably assist archival processes, or to make them possible in the first place. Consequently, its objectives range from ensuring the appropriate creation and disposal of record entities, to identification and authentication of collection items, rights management, and documenting the history of a record.
Between these two aspects of metadata transitions are fluent, overlappings do exist. An attempt to find common ground between the various formats trying to reconcile the differing backgrounds where metadata is used was made, most notably, by the definition of the Dublin Core Metadata Element Set8. In contrast to other strategies that try to incorporate all possible requirements obtaining very packed, almost overloaded format sets, Dublin Core identifies a small, simple set of metadata elements designed to be applicable to any communities needs. Its elements build on the principles to guarantee extensibility, allowing augmenting the core set of elements with more specialised data, and to allow for maximum flexibility, keeping the elements optional to use and repeatable, for no respectively multiple entries. Originally designed to aid resource discovery in the Web, the Dublin Core Metadata Set is now being used for a wide range of specialised areas for multiple purposes, not only with digital but even with real, physical objects [WK00]. Due to its simple and flexible approach, as well as the consensus of various communities on its basic structure, it offers the possibility of semantic interoperability across metadata formats in different disciplines.