RE: What to do about namespace derived URI refs... (long)

> -----Original Message-----
> From: ext Jonathan Borden [mailto:jborden@mediaone.net]
> Sent: 06 June, 2001 19:04
> To: Patrick.Stickler@nokia.com; www-rdf-interest@w3.org
> Cc: Ora.Lassila@nokia.com
> Subject: Re: What to do about namespace derived URI refs... (long)
> 
> 
> Patrick,
> 
> 
> >
> > Likewise, if I wish to define enumerations of token values for
> > a property in XML Schema and then assign rankings (numeric
> > values) for those in RDF Schema, the URI ref I have to use
> > is tied to the XML Schema MIME content type and is not generic
> > and portable, as it should be.
> >
> >    XML Enumeration for 'status'         RDF assigned rank
> >    ---------------------------          -----------------
> >            draft                                1
> >         draft_approved                          2
> >          approved                               3
> >           retired                               4
> >
> > (the rank is needed if I want to have queries such as
> >  'status <= approved', etc.)
> >
> > Tricky, yes. Though I'd use a stronger term ;-)
> >
> 
> It would be great to have a MIME media type independent 
> syntax for fragment
> identifiers particularly for RDF and like applications 
> (otherwise we really
> are kidding ourselves assigning semantics to fragment 
> identifiers -- well at
> least IMHO). 

This is my point entirely. A URI ref is just a URI ref. It points
to some real location within a real MIME data stream. It is not
in-and-of-itself an identifier suitable for assigning useful
semantics (I stress the word 'useful' ;-)

E.g I have an ontology which defines numerous properties, each
of which have several classificatory properties (properties
of properties) and many of which have controlled value sets,
where each value in that set also can have properties defined
for it (rank, label, class membership, etc.).

This ontology defines an abstract semantics, not a serialization.

It could be serialized in many ways, with that serialization
defined and checked by any number of schemas -- yet even with
dozens of different serializations and even more schemas, it
is the *same* ontology and hence in all cases should result
in a single consistent representation in an RDF knowledge base.
The only way this can happen is if (a) there is a mechanism
for defining that abstract ontology irrespective of serialization
(e.g. some URN scheme) and (b) there is an explicit and 
consistent mapping from every serialization/schema context
to that single identity scheme.

The power and benefit of RDF is not in its serialization mechanisms.
They are trivial. It is in the data model that defines the
knowledge space within which  the semantic web will operate. There
must be a reliable many-to-one mapping from all serializations
to that knowlege space which maintains equivalent and consistent
identity for common resources, abstract or concrete.

At present, as far as I can see, there is none.

The scope of serialization formats and serialization schemas such
as DTDs, XML Schema, etc. is a single unit of syndication into
a larger knowledge space. The scope of RDF Schema, DAML, etc. is
that fully syndicated knowledge space. The serialization is a
trivial necessity for information storage and interchange, insofar
as the semantic web is concerned. What is central are the ontologies
and the consistent, global, standardized, and explicit identity
of their components.

XML Schema, nor any MIME content type, should not, cannot dictate
the unique, portable, consistent identity of abstract concepts
on the semantic web.

It is wonderful that XML Schema has a rich fragment identification
scheme, and I applaud the work done there -- but however rich and
wonderful it might be, it cannot solve the problem of global,
universal identity of components of abstract vocabularies, taxonomies and
ontologies.

> For example: http://www.rddl.org/fragment-syntax

I'll have a look. Thanks.

Cheers,

Patrick

Received on Thursday, 7 June 2001 03:10:46 UTC