On Tue, 2005-03-08 at 23:21 +0000, Pete Johnston wrote:
> On Tue, 8 Mar 2005 21:50:03 +0000, Andy Powell <[log in to unmask]> wrote:
> >As you say, the RDF encoding of DC can integrate seemlessly with the RDF
> >encoding of SKOS... but an SKOS concept scheme could *not* be used in
> >place of a DC vocabulary encoding scheme in our XHTML or XML syntaxes.
> True, but that's a limitation of those syntaxes (specifically their
> lack of support for multiple descriptions).
> I feel a bit uneasy about arguing that the designers of other
> ontologies need to consider DC's modelling "style" when making their
> modelling decisions; I'd definitely feel uneasy making that argument
> on the basis of the limitations of two formats that we already
> acknowledge don't implement the DCAM fully.

Well, I think it's somewhat inevitable. What if DC used some other
method of linking Vocabulary encoding schemes to a value, such as a
dcterms:inScheme property? That wouldn't help a bit, as ontology
designers would still need to be aware of this property in order to be
compatible. Compatibility isn't something that happens when you look the
other way... and now that DC and SKOS *are* aware of each other,
compatibility is at least an option.

That said, it isn't really a question of incompatibility, just
incompleteness. The SKOS core guide says:

  "To use SKOS-Core with Dublin Core, make concepts the objects of
   statements using the dc:subject property."

There is no mention of vocabulary encoding schemes. And this will work
in all syntaxes that support using value URIs.

What would be needed to complement SKOS is a notion of "class of
concepts". As SKOS has chosen not to introduce that, there simply are no
vocabulary encoding schemes for SKOS values. Or if you like, you can add
one of your own without breaking anything. Incomplete, but not wrong.

Of course, it would be nice if SKOS and DC could sort it out. It would
be a pity if SKOS would fall into the class of interoperability
standards that tries to do things afresh without looking at existing
standards and practices. While using rdf:type is certainly a modeling
choice by DC, I don't think it's entirely unreasonable.