RE: web proper names

> -----Original Message-----
> From: ext Hamish Harvey [mailto:hamish@floodrisknet.org.uk]
> Sent: 22 September, 2004 13:00
> To: Stickler Patrick (Nokia-TP-MSW/Tampere); JohnBlack@deltek.com;
> henry.story@bblfish.net; RDFInterest
> Cc: h.halpin@ed.ac.uk; ht@inf.ed.ac.uk
> Subject: RE: web proper names
> 
> 
> 
> On Wed, 22 Sep 2004 11:38:52 +0300, Patrick.Stickler@nokia.com said:
> 
> > With an approach such as URIQA, though, the degree of 
> standardization
> > and concensus reflected in e.g. programming languages, document
> > models, ontologies, etc. is not required.
> 
> Standardisation of programming languages is rather a different beast
> from standardisation of "ontologies", I think. Although the 
> symbols used
> in a programming language are generally adopted from a 
> natural language
> (usually English), they are defined in terms of what they do. The
> difficulty here is in describing this unambiguously in English prose,
> but the symbols themselves are firmly grounded in reference
> compilers/interpreters.
> 
> Standardising an schema/ontology is a whole different matter. The (IMO
> unfortunate) overloading of <ontology> to mean the expression in a KR
> language of what used to be called an ontology (approx. an account of
> what is) points to the problem. The terms in a schema/ontology can be
> grounded in a real ontology, which can only be imperfectly shared
> between individuals, and probably only imperfectly understood by a
> single individual.

Clearly, two agents trying to communicate will have to have
some mutual vocabulary of terms in common. Though it may be
that until they communicate, they won't know which terms they
both understand.

My point is that, with URIQA, there is much greater freedom
to use vocabularies that have primary local application but
which are, by RDF statements, related to other, more commonly
used, industry standard vocabularies, such that if some agent
encounters one of the more localized terms, and asks about it,
it may then find out that it is e.g. rdfs:subPropertyOf or
rdfs:subClassOf some term that it knows the meaning of, and
can then utilize statements employing that more local term
effectively. Thus, because of the power of RDFS and OWL, agents
can in essence identify shared semantics of terms on-the-fly,
and URIQA enables them to obtain the information to do that
based solely on the URI alone.

True, without some number of core, foundational vocabularies
and ontologies, to which more specialized or localized vocabularies
are related, arbitrary agents will be hard pressed to interoperate,
but my key point here is that with a critical mass of deployed
URIQA-enlightened servers, there is far less of a burden to
standardize on a minimally few vocabularies/ontologies which must
have native support by local systems.

> 
> > And insofar as URIs are concerned, approaches such as URIQA
> > specifically address the issue of "what does this URI mean?"
> > in a formal, machine understandable manner.
> 
> Ummm, I like URIQA, but I'm not convinced it does this at 
> all. A CBD can
> be an answer the question "what does this URI mean?" IFF what 
> it "means"
> is a sequence of bytes which can be retrieved (and this is 
> unambiguously
> specified in an agreed RDF vocabulary). Otherwise all it can do is
> provide a set of properties which (help to) establish identity of some
> still inaccessible-to-the-machine entity. If URIQA is supposed to
> provide meaning beyond identity it sounds like an entrance way to the
> hermeneutic hall of mirrors [1]. 

I was speaking colloquially. Let me put it a different way: URIQA
provides efficient access to authoritative descriptions of resources
which are expressed in a formal manner conformant with the RDF MT.

Is that better?

Cheers,

Patrick

Received on Wednesday, 22 September 2004 10:29:26 UTC