On Fri, 2005-03-11 at 17:23 +0000, Miles, AJ (Alistair) wrote: > > But won't most people do e.g. > > <http://somewhere/somedoc> dc:subject <http://somewhere/someconcept> . > > ... ? I.e. in this (most common?) use case the user is referring > directly to a conceptual resource via the URI of that resource, and so > never has to refer a vocabulary encoding scheme. Can you outline the > usage scenarios involving dc:subject that require a reference to a > vocabulary encoding scheme? It's true that the URI is a unique reference to the concept, even without reference to a VES. However, there are a number of scenarios where an explicit VES is helpful: * When converting between different encodings (XML and RDF, typically). * When presenting/editing metadata records and the presentation/editing software is vocabulary-aware, a VES will help present/select between concepts. See for example our work on SHAME [1]. * Dumb-down [2] requires presence of at least rdfs:label in order to work. If you copy the rdfs:label of your conceptual resource into the instance metadata, it's reasonable to copy at least rdf:type too. It's true that all scenarios assume that for some reason, the full RDF for the concept is not available. Providing duplicate data fragments is a kind of service to DC consumers, not an absolute requirement. [1] http://kmr.nada.kth.se/shame/ [2] http://dublincore.org/documents/abstract-model/#sect-5 > Aaaah you used the 'T' word ('term')!!! I would do something like: Oh, sorry ;-) I guess it's a *very* good idea to avoid usage of that term (oups, I mean *word*!). Especially in a spec such as SKOS. But people must be allowed to use it in their own thesauri, right? I mean, if the organization behind AAT uses the word "term" in their document, it's only reasonable to use it in the RDF as well...? But not in SKOS, I agree. /Mikael