I want to add that this approach (URI-encode all vocabularies) is what
we have recommended for the IEEE LOM RDF binding. It also allows you to
describe the hierarchy of your taxonomy in a separate RDF model, adding
translations of terms in others, etc. It's a very nice approach,
naturally.
So I wholeheartedly agree with your analysis of the two rdf:value uses -
none of them are strictly necessary any longer. Should DCQ/RDF change to
reflect this? I don't know. In a later version, most certainly! Right
now no RDF tools support datatypes, however...
/Mikael
On Fri, 2002-12-06 at 17:05, Patrick Stickler wrote:
> > _:a dc:subject _:x .
> > _:x rdf:type dcterms:MESH .
> > _:x rdf:value "D08.586.682.075.400" .
> > _:x rdfs:label "Formate Dehydrogenases" .
>
> Hi folks,
>
> A question...
>
> Wouldn't it be more efficient and optimal to use URIs for
> cases such as above, which would (a) eliminate the bnode
> reducing four triples to one and (b) define the label once
> for all instances of a given value, and (c) allow one to
> use generic RDF reasoning to determine equality of values
> rather than having equality depending on DC-specific
> intepretation of the above idiom. E.g.
>
> _:a dc:subject <http://purl.org/dc/terms/MESH/D08.586.682.075.400> .
> <http://purl.org/dc/terms/MESH/D08.586.682.075.400> rdfs:label "Formate Dehydrogenases" .
>
> or more compactly
>
> @prefix mesh: <http://purl.org/dc/terms/MESH/> .
> _:a dc:subject mesh:D08.586.682.075.400 .
> mesh:D08.586.682.075.400 rdfs:label "Formate Dehydrogenases" .
>
> It seems to me that there are two primary uses of rdf:value
> in DC, one for datatyped values and another for controlled
> code sets. Typed literals in the new RDF specs handles the
> first case, and a URI based approach such as above would more
> optimally handle the second (since the values in question are
> resources about which additional statements must be made,
> such as label).
>
> One metric I've employed for some time is that, if a value needs
> a label, or is a member of a controlled set, use a URI to denote
> the value.
>
> Given the fact that rdf:value has no semantics defined in the MT,
> and is, while not deprecated, also has no normative interpretation
> defined, opting for typed literals for datatype values and
> URIs for controlled termset values seems a better way to go.
>
> Eh?
>
> Just thinking out loud... ;-)
>
> Patrick
>
>
> [Patrick Stickler, Nokia/Finland, (+358 40) 801 9690, [log in to unmask]]
--
Plus ça change, plus c'est la même chose
|