 |
 |
Carl, I had a quick look at your paper today. More detailed comments will follow , but my first reaction was that it lacks an overview to motivate the concepts that follow. Here is my attempt at an introduction. Please feel free to extract anything that is useful. Bill ======= Overview In March 1995, OCLC hosted a meeting to discuss metadata for items of digital information. The major result of that meeting was a list of thirteen metadata elements that can describe a wide variety of items. This list has become known as "the Dublin Core". A year later a follow up meeting was held at the University of Warwick to review progress and plan future steps. Three key concepts came out of this meeting. Collectively, they have been nicknamed "the Warwick Framework." Metadata Packages Although many groups are building information services with metadata drawn from the Dublin Core, every group is adding extra metadata elements. The additions may be subject specific (e.g., for geo-spatial data), technical (e.g., formats or protocols), structural (e.g., links to show relationships between complex objects), or business related (e.g., terms and conditions for usage). To handle this need, the Warwick meeting proposes a set of metadata packages. For example, the Dublin core is one package; another might be the terms and conditions package. An information service can select one or more packages to provide metadata for a set of objects. This approach has several advantages over selecting individual metadata elements from a very long list of elements. Packages can be very different. For example, a package that expresses relationships among objects might use abstract data structures. A reasonably small list of well defined packages is hoped to enhance interoperation and lead towards standardization of practices. In addition, as described below, packages allow flexibility in the development of a security architecture. Security When a digital object is accessed over a network, there are many occasions when a supplier wishes to make only part of the metadata accessible to specific users. For example, an organization may need to have access to technical metadata in order to store and transmit information, but, to avoid potential liability, may explicitly desire not to have access to metadata that describes content. A commercial organization may wish to provide some metadata openly, but require authorization before giving access to other metadata. These objectives can be achieved by providing each metadata package with its own security. Access controls on each package can be different. Representation of Metadata There will undoubtedly be many different representations of metadata within repositories. For example, the metadata for a digital item can be embedded within the item or external but associated. Much of the work on repositories uses the concept of a "digital object", in which the metadata and the data are both stored within a repository without the details of the storage mechanism being known externally. Formats for exchanging metadata between systems need to be clearly defined, flexible, yet easy to use. Preliminary work carried out during the Warwick meeting convinced many of the people attending that SGML provides a suitable format to represent metadata packages. The meeting considered that Web pages in html format are such an important special case that they deserve special attention. The meeting proposes a syntax based on the html "meta" tag.