On Mon, 2 Jul 2001, Roland Schwaenzl wrote:
> In Math-Net we use rdf:Bag in the dc:creator field for collaborative efforts.
> rdf:Alt is used in dc:identifier.
>
> Both uses fit with the usage described in RDF M&S. We currently don't use rdf:Seq.
M&S does encourage this, yes. However I have doubts about this as a
modelling style that will scale:
Does dc:creator ( doc1, bag1)
rdf:_1 ( bag1, dan )
rdf:_2 ( bag1, eric )
...imply
dc:creator ( doc1, dan)
and
dc:creator ( doc1, eric) ?
Not in the general case: RDF doesn't allow us to make this inference. Same
for Seqs and Alts. However in Dublin Core, we want to be clear that we
believe both dan and eric to be a creator of the document. The thing of
rdf:type rdf:Bag didn't do the creating, after all. If we believe that the
members of a collective authoring team aren't individually creators, I'd
argue that dc:contributor is more applicable. In the example above, by
using rdf containers, we risk obscuring the key pieces of information:
that both dan and eric stand in a dc:creator relation to doc1.
My reading is that it is down to the creators of specific RDF properties
(foo:author, dc:creator) to decide whether it makes sense for them to
point to containers, and what it means when that happens.
For example, if I create an RDF property util:foo and I say it has an
rdfs:range of some class util:Bar, it would be a mistake to point util:foo
at an rdf:Seq that was a list of things of type util:Bar. Containers
aren't supported in that way in RDF, they're just a take-it-or-leave-it
utility vocabulary which may make sense for some applications to use.
As far as 'Alt' for identifiers, doesn't the fact that several things are
represented as being the dc:identifier already establish that there are
multiple alternative identifiers? I can't see what rdf:Alt adds in that
case.
Dan
|