W3C home > Mailing lists > Public > www-webont-wg@w3.org > September 2002

Re: possible semantic bugs concerning domain and range

From: Christopher Welty <welty@us.ibm.com>
Date: Wed, 25 Sep 2002 11:20:34 -0400
To: www-webont-wg@w3.org
Message-ID: <OF41873928.47114A35-ON85256C3F.004B11AE@pok.ibm.com>

In trying to peice this debate together, I find myself more and more 
puzzled by Pat's position.

It seems to have come out that Pat has some special status in mind for the 
notion of a "range". That is, that there are special classes which 
explicitly denote ranges of properties - in fact, I gather (although I 
haven't found it expressed formally) that Pat wants there to be something 
like a class for each property which denotes "THE range" for that 
property.

The debate started out with a fairly simply example of what Peter felt to 
be a missing entailment, i.e. 

Given (1):

        foo rdfs:range baz
        baz rdfs:subClassOf bar

The following holds in every model:

        _y foo _x
        _x rdfs:Type bar

And therefore:
        foo rdfs:range bar  (2)
is consistent with any model of (1).    Surely Pat does not dispute this. 

Pat's issue, as I understood it, was whether the triple (2) should be 
explicitly entailed in the RDF(S) encoding of OWL, since it seems rather 
trivial and arbitrary - there are a huge number of sentences that are 
trivially true given a set of initial sentences, why should we require 
that certain of them be explitly entailed?

Entailments, as Pat well knows, can be useful in computational systems and 
often decisions about which entailments to include and which to ignore are 
made for computational reasons.   In this case, I think there are useful 
things that can follow from (2):

If we add the following to (1):

        oof rdfs:subPropertyOf foo (3)
        oof rdfs:range rab (4)

Then the RDF graph of (1 + 3 + 4) is consistent.  However, if we add:

        rab owl:disjointWith bar  (5)

This turns the graph (1 + 3 + 4 + 5) into an inconsistent one.  It seems 
to me difficult to find, computationally, this inconsistency if you don't 
have (2), and quite simple to find if you do.    Note that when I say 
"find this inconsistency" I mean to compute it given only the graph 
(1+3+4+5) and nothing else, in particular no instances of the classes or 
properties, just the entailments.

I'm basing this "difficult" and "easy" to find judgement on the fact that 
I believe:

        (1+3)  |=  oof rdfs:range baz (6)
 
And thus that (6 + 4 + 5) together are inconsistent, but to prove this 
requires a lot more work unless you have (2).

What do you think?


-Chris
Received on Wednesday, 25 September 2002 11:21:08 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:52 GMT