Re: 5.20, need for synonyms

Peter F. Patel-Schneider wrote:

>
> > For example, rdf:Property becomes an instance of rdfs:Class in
> > > RDFS.
> >
> > I've not seen a good explanation for why one is in "rdf" and the other
in
> > "rdfs" ... I'd not like to propogate mistakes.
>
> And how are we supposed to fix this mistake, if mistake it is?

To clarify, what I consider a mistake is not that rdf:Property becomes an
instance of rdfs:Class, rather that the related concepts "rdf:Property" and
"rdfs:Class" are defined in different namespaces. It causes endless
confusion among developers ... how to remember that its _rdf_:Property
rather than _rdfs_:Property ? How do people keep track of this? I still have
a hard time remembering which is which.

>
> Well daml:Class is special in that it has a different meaning from
> rdfs:Class, so I would say that if one is interested in owl:Class then no.

This is the heart of the issue. If owl:Class has a different meaning than
rdfs:Class we need to refer to it by its OWL namespace qualified name:
owl:Class i.e. it would not be acceptable to change the definition of
rdfs:Class when using OWL, for example.

...

>
> > C: Would it be possible to refer anyone seeking a definition of an OWL
Class
> > to the RDFS documentation?
>
> Again, no, because if one cares, daml:Class (and thus probably owl:Class)
> is different from rdfs:Class.

If so, then by using the qualified name owl:Class, the namespace URI ought
reference some documentation that tells us about "Class"

> >
> > Just so. When encountering a QName, one ought be able to dereference the
> > namespace URI and retrieve a document into which one can find a
definition
> > by using the QName local-name as an identifier e.g.
>
> How can this be done?  RDF does not have namespaces or localnames, just
URI
> references.  See the recent discussion in the RDF Core WG on this.

Sure, but in terms of dereferencing a URI reference, this is all worked out
quite well. A user-agent/browser sends only the URI part (everything before
the '#") to the server, and gets back a document which is a representation
of the resource. The user-agent then uses the fragment-id (everything after
the '#") to locate a part of the document which contains the definition (or
some information about) what we are talking about.

So in general -- and in specific when our namespace URI ends in '#" as it
does --

QName prefix => URI
QName local-name => fragment-id

and this syntactic distinction is encoded in a URIreference by the '#'

>
> Further, there is no notion of a definition in RDF, just statements.
> Similarly OWL has no notion of a definition.  What is a definition?  How
> will you be able to find (all of) it?  The very idea of a definition seems
> to go against the RDF vision.

I'm just talking in plain English here, it might be what a developer looks
for. It also might be something that has an rdf:ID attached to it. In RDF
just as in English, although one is not constrained to define something in
one place, one _can_ and good practice suggests good organization of
documentation. We _can_ do it, even though are languages do not mandate it.

In any case I don't think there is a _need_ for synonyms, if we are talking
about true synonyms that are really equivalent. On the other hand when we
have real differences e.g. with <rdfs:Class> vs. <owl:Class> and OWL has a
need to discuss such differences, or give its own definition for something.,
then we should put something in the OWL namespace.

Jonathan

Received on Saturday, 27 July 2002 09:19:38 UTC