W3C home > Mailing lists > Public > www-rdf-interest@w3.org > November 2000

Re: A triple is not unique.

From: Dan Brickley <Daniel.Brickley@bristol.ac.uk>
Date: Mon, 20 Nov 2000 19:12:38 +0000 (GMT)
To: Jonathan Borden <jborden@mediaone.net>
cc: Dan Brickley <Daniel.Brickley@bristol.ac.uk>, "McBride, Brian" <bwm@hplb.hpl.hp.com>, Seth Russell <seth@robustai.net>, RDF-IG <www-rdf-interest@w3.org>
Message-ID: <Pine.GSO.4.21.0011201903410.4324-100000@mail.ilrt.bris.ac.uk>

hi

On Mon, 20 Nov 2000, Jonathan Borden wrote:

> Dan Brickley wrote:
> 
> > On Mon, 20 Nov 2000, McBride, Brian wrote:
> > > I can see that the disruption caused to current implementations
> > > would be a factor if m&s was ambiguous.  But if the answer lies
> > > in m&s, I humbly suggest the spec takes precedence.  That's what
> > > specs are for.
> ...
> >
> > Maybe I missed the appropriate post, but I'm unclear how we square the
> > set-oriented definition of 'Statement' with the syntactic ability to
> > assign various IDs (and hence URIs) to the XML occurances of RDF
> > statements. Or rather, I suspect we could do this, by adopting a strong
> > view on the "can a resource have multiple URIs" question that
> > periodically bedevils discussion here and elsewhere.
> 
> This is looking more and more like a gaping hole growing between the model
> and syntax parts of the specification.

Perhaps. If forced to choose, I have little doubt that most implementors
would take the model as primary. That said, I still lean towards the
'multiple statements per p/s/o' policy, since it fits best with most of
the use cases I can think of for RDF reification.

> 1) Just because it is syntactically possible to assign an rdf:Statement an
> ID doesn't mean that it ought be allowable to assign more than one ID to the
> same statement.

How would you propose to police that? rdf:Statements will be manifested
in thousands of RDF documents scattered around the Web. Does a statement
gets its ID from the first (?!?) person to write it down in XML?

> 2) I strongly caution against trying to wrangle out of this issue using the
> "can a resource have multiple URIs" question which rears its head from time
> to time.

I agree. Discussing this with Brian and others in Bristol last week, we
hit upon that as one of the various possible exit strategies for this
confusing topic.

	 Using the RFC 2364 definition of URI it is clear that the resource
> identified by a URI may be abstract and hence *even when 2 URIs resolve to
> the same network entity*, each URI still identifies a distinct abstract
> resource. The distinction between the resource identified by a URI and an
> entity retrieved when a URI is resolved is clearcut.
> 
> This view has already been definitively stated in RFC 2364 however if Tim BL
> himself wishes to update the definition of a URI in order to solve this
> apparent problem in the RDF rec I am all ears. In the meantime this issue
> can be laid to rest by clarifying the uniqueness of a statement as defined
> by (p,s,o)

Nobody suggested TimBL was going to any such thing. I suspect he has the
opposite view, FWIW. I mentioned that topic as one that has seemed (to
some at least) as not yet open-and-shut, and one that would offer us a
trick for allowing a single p/s/o triple to identity the selfsame
rdf:Statement resource even if it had multiple URI identifiers. This, to
me, doesn't seem a good route, but I wanted to mention it for
completeness.

Dan
Received on Monday, 20 November 2000 14:12:53 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:51:46 GMT