 |
 |
Dear all, I'd like to pick up on two important threads from discussion over the summer. In this message, DCAM. On Fri, Aug 07, 2009 at 12:03:35AM +0200, Mikael Nilsson wrote: > I have been sketching on a replacement of the current DCAM, substituting > a much leaner version directly building on the RDF model while retaining > the same external interface. I've called this DCAM 2.0, and the latest > sketch is here: > > http://dublincore.org/architecturewiki/DCAM-2.0 Mikael has sometimes characterized this as a "re-factoring" of DCAM that keeps the same interface to the Description Set Profile draft, but de-emphasizes DC-specific terminology wherever possible, in favor of simply pointing off to RDF directly. As I see it, DCAM has evolved the way it has for historical reasons: -- Roughly 2002 to 2005 was the formative period of DCAM, leading to the DCMI Recommendation of January 2005. The Dublin Core community had by then developed a "grammar" of elements, element refinements, and encoding schemes, but this grammar lacked a formal modeling basis. During this time, RDF was finalized as a W3C Recommendation (in February 2004), but RDF was seen by many in the DCMI community as a research framework with an overly complicated XML syntax. The 2005 DCAM moved the Dublin Core model alot closer to RDF in modeling terms (e.g., by differentiating between Vocabulary Encoding Schemes and Syntax Encoding Schemes), though it expressed the model primarily in the terminology of the earlier "grammar". The relationship of DCAM to RDF was elaborated in an appendix as one of several alternative encoding syntaxes for the model. -- The 2007 revision of DCAM made the relationship to RDF more explicit, for example by equating the Syntax Encoding Scheme with the RDF datatype. This led to a complete review of DCMI metadata terms with regard to formal domains and ranges, type of encoding scheme, and the like, with a major revision of the terms published in January 2008. Since DCAM 2007, I would argue, the Linked Data movement and the rapid success of "structured search" using RDFa have made RDF both more familiar and less controversial, so the time seems ripe for DCAM to be "re-factored" more explicitly on an RDF basis. This would result in a visibly a _shorter_ DCAM, because entire sections (e.g., "DCMI Resource Model" and "DCMI Vocabulary Model") would be dropped. It would mean a clearer relationship to RDF, because DCAM 2.0 would more clearly highlight the few really DC-specific (i.e., non-RDF) things about DCAM -- the "description" and "description set" constructs, the notion of a vocabulary encoding scheme, and the DCAM "statement" construct, in effect a community-specific pattern for RDF graphs (using, where needed, Vocabulary Encoding Schemes or multiple value strings in parallel). The existing document "Expressing Dublin Core metadata using the Resource Description Framework (RDF)" could become redundant. Work can proceed on Description Set Profile language without such a re-write of DCAM, but I believe such a new version would make that work easier -- and easier to explain. Tom -- Thomas Baker <[log in to unmask]>