Re: Meaning of URIRefs

From: Sandro Hawke <sandro@w3.org>
Subject: Re: Meaning of URIRefs 
Date: Fri, 25 Oct 2002 14:51:06 -0400

> 
> > If I understand Sandro's position correctly, then if I use a URI
> > foo:bar (when oh when are we going to finally add qnames to RDF?),
> > I'm committed to "believing" all of the information
> > stored in the corresponding document.
> 
> Right.   
> 
> I thought I'd addressed your two points in my reply to Peter; I'll try
> again.
> 
> > Below, Sandro
> > tempers that by saying you agree to all of the 'definitional'
> > information.  There are at least two problems with Sandro's
> > proposal.  One is that long experience with KR and DL systems
> > indicates that a line between 'definitional' and 'non-definitional'
> > is completely arbitrary, and therefore untenable as the basis
> > of an RDF convention.  For example, I would hope that we
> > would all agree(?) that an rdf:type statement is 'definitional' (how
> > else do we define constants like PI?).  The statement
> >       (foo:GeorgeBush rdf:type foo2:Person)
> >   is perhaps non-controversial.  However the statement
> >      (foo:GeorgeBush rdf:type foo2:SittingUSPresident)
> > is not definitional.  A fanatical Gore supporter might claim that
> > Al Gore is still the rightful US President.
> > 
> > Continuing, suppose the retort is that 'rdf:type' is NOT a definitional
> > statement.  Then we have a disagreement over what should be
> > definitional and what should not, which just backs up my claim above.
> 
> My proposed split-off of definitional information is NOT intended to
> be done mechanically.  I tend to agree with your point that  dividing
> "definitional" and "non-definitional" is, in the end, arbitrary.
> 
> I propose, rather, that the split be done by authors and publishers.
> If you are "coining" a URIRef which you want others to use, you should
> provide good definitional content, not contaminated by any other
> information.  You should perhaps also make strong commitments to
> steward that definition, but that's a different issue. 

The point of Bob's message, I think is that the only safe place to draw the
line is at

       (foo:GeorgeBush rdf:type rdfs:Resource)

Any other drawing of the line can be objected to.

I strongly agree with this reasoning.

> > A second problem is that I might want to disagree with a statement
> > by citing the URI, and Sandro's proposal appears to prevent that.
> > If I want to state that the individual denoted by 'foo:GeorgeBush'
> > is not the president, is not a Person, or whatever, I'm going to have
> > to reference 'foo:GeorgeBush' in my refutation.
> > 
> > I imagine that Sandro might like to amend his proposal to allow
> > quoted references to URI's without committing to their definitions?
> > If we had quotation.  Otherwise, it would be hard to express (in RDF)
> > a disagreement with how someone else had constructed their ontology.
> 
> I don't have to amend my proposal to allow this, since nothing I said
> prevents it; you just have to keep such quoted URIs inside string
> literals rather than use them as node or arc labels on your RDF graph.
> 
> But, again, there is a way to point to something without using a
> quoted URI or a URI at all: you can refer to it by some collection of
> properties which identify it unambiguously.  This is a technique often
> used in RDF for identifying people.   It requires more difficult
> reasoning, but if for some reason you can't obtain or create good
> definitional content, it works.

This proposal has even more problems.  

How can I talk about George-Bush-the-lesser without commiting to some
ancillary facts about him, such as the claim that he is the 43rd president
of the United States?  Even worse, I have to commit to some well-known
notion of president and United States, which I may be unwilling to do.
(I'll let the use of 43 go by, for now at least.)

>     -- sandro


Peter F. Patel-Schneider
Bell Labs Research

Received on Friday, 25 October 2002 15:52:22 UTC