5.20, need for synonyms

Let me state my position for the archives:

If any concept, token, QName or URIreference is defined _exactly_ as in
RDF/RDFS it is correct to use the rdf/rdfs namespace.

If any concept/token/QName or URIreference is defined _in any small or even
trivial_ way differently for OWL than RDF/RDFS, then this concept should be
given a name in the OWL namespace.

My strong view is that one should look to some document referenced by the
namespace URI, to find definitive documentation about a QName.

For example:

rdfs:Class vs. owl:Class

if these two are _exactly_ the same, which means that the definition of
owl:Class is entirely given by the RDF MT, and that the OWL MT adds nothing,
or reinterprets nothing, then there is no need to import the local name
"Class" into the OWL namespace.

same for rdf:Property etc...

So any concept defined within the OWL MT should have a corresponding QName
whose namespace is the OWL namespace -- and the OWL namespace URI should
reference some definative OWL documentation -- preferably in OWL ooops, no I
mean RDDL :-))

My initial preference which I'd expressed in the conf call was to liberally
import names into the OWL namespace (ala DAML+OIL) but DanC has good
arguments against that. I'm not opposed to importing RDF names into OWL (as
long as proper owl:sameXAs's are defined) _however_ I really don't care if
we use:

rdf:label rather than owl:label

Again, if the OWL MT 'reinterprets' or extends the interpretation of some
QName/URI then the OWL namespace MUST be used.

Food for thought:

<owl:oneOf>
    <rdfs:Class rdf:ID="foo" />
    <rdfs:Class rdf:ID="bar" />
</owl:oneOf>

vs.

<owl:oneOf>
    <owl:Class rdf:ID="foo" />
    <owl:Class rdf:ID="bar" />
</owl:oneOf>

That depends on whether rdfs:Class isEquivalentTo owl:Class in every
respect, or if not, whether owl:oneOf cares...

Jonathan

Received on Friday, 26 July 2002 16:11:38 UTC