On Mon, Feb 16, 2015 at 01:03:59PM +0100, Antoine Isaac wrote:
> Because the former is too close to existing OWL modeling. The
> rationale for distinction between
> ldom:property/ldom:predicate/ldom:minCount and the 'regular' OWL
> constructs is much more difficult to get when either set can be
> attached to the same kind of resources, using patterns that furiously
> look the same. I expect some people will remain confused for a long
> time. And may want to directly attach shapes constructs to the 'real
> classes' from the ontologies they (re-)use.
+1 to avoid patterns that furiously look the same
I also wonder if there is an implicit assumption here that RDF data will
be grounded in RDFS or OWL classes. Cow X can be described with:
Cow X
http://purl.org/dc/elements/1.1/type
http://aims.fao.org/aos/agrovoc/c_1939
where agrovoc:c_1939 is the SKOS concept in AGROVOC for 'cows' and the
dc:type property has kinda the same meaning as rdf:type but without the
RDF class semantics.
RDF is only fifteen years old and still evolving. Why limit an RDF data
pattern language to data patterns structured around RDF _classes_? I
would venture to guess that the number of available SKOS concepts, for
example, is increasing much faster than the number of RDF or OWL
classes. Or should people be encouraged to say 'CowX _rdf:type_
agrovoc:c_1939'? Or should they be told that SKOS concepts are not
usable to say what something is... even though SKOS concepts can look
furiously the same as RDF classes...?
Tom
--
Tom Baker <[log in to unmask]>
|