 |
 |
There are two descriptions given of rdfs:isDefinedBy in <http://www.w3.org/TR/rdf-schema/>. The first has already been quoted on this list: "The property rdfs:isDefinedBy is a subproperty of rdfs:seeAlso, and indicates the resource defining the subject resource." The second appears in the tabular summary of the properties (predicates) of RDF and RDFS: "Indicates the namespace of a resource". It's probably also worth adding the comment given in the RDFS for RDFS: "Indicates a resource containing and defining the subject resource." The second of these is the most precise (that is in using the term "namespace" it references a subset of "resource" used elsewhere). The question is, is this a mistake? and is this normative? It is in a summary, but there is nothing to indicate it isn't normative. This would seem to indicate that the defining resource is a namespace. My reading so far has been that the defining resource isn't necessarily a namespace (because there is nothing in RDF to prohibit URIs that can't be built from namespace URIs), but it is a namespace for those resources used in RDF/XML (which can only use URIs that can be built from namespace URIs), including all resources defined by Dublin Core. Certainly it almost definitely *does not* reference an RDFS document as such. To define the resource at the URI <http://purl.org/dc/terms/> as identical to the document obtained by dereferencing <http://purl.org/dc/terms/> would break the general understanding of what a web resource is, and can only work if you somehow assume that dereferencing that URI would always return the same kind of document. It would also cause problems with the unarguable fact that <http://purl.org/dc/terms/> is a namespace. Actually that last point is the clincher as far as I'm concerned. <http://purl.org/dc/terms/> is a namespace, and unless you can convince me that it is not, you can not convince me that the object in the statement <dcterms:alternative> <rdfs:isDefinedBy> <http://purl.org/dc/terms/> . is not a namespace. So really we are left with the hoary "namespace document" debate. The RDFS document[1] obtained by dereferencing <http://purl.org/dc/terms/> is a namespace document. It is certainly not something that would match many people's arguments about what a namespace document *should* be, but that doesn't stop it being a namespace document. The RDFS document one obtains from dereferencing the URI is a representation of the namespace. One could also obtain an n3 document, a HTML page documenting the namespace, and RDDL file, or a pretty picture. If the server were set up to return a HTML page when asked for text/html or application/xhtml+xml and an RDFS document when asked for application/rdf+xml for the same URI that wouldn't change the resource the URI identified, only the representation obtained. [1]Actually it's a stretch to say there is a RDFS document obtained from <http://purl.org/dc/terms/> at all, since the server assures me that it's actually a text file, but presumably that bug with the headers will be fixed and it is *meant* to be an RDFS document.