RE: Could owl:sameAs reference non-OWL resources

Re: Could owl:sameAs reference non-OWL resourcesDear Pat

Your response is clear and satisfactory, and sums up clearly a few exchanges
I had about it since the question was set, through which, I hope, my
understanding of the debate about URIs as names vs URIRefs linked to a
particular addressing protocol has clarified a bit. It figures interaction
between OWL and Subject Indicators might turn out to be less obvious than
what I imagined a few months ago. But I still hope to come out with some
proposal about it.

Thanks for your time

Bernard Vatant
Senior Consultant
Knowledge Engineering
Mondeca - www.mondeca.com
bernard.vatant@mondeca.com


 -----Message d'origine-----
De : public-webont-comments-request@w3.org
[mailto:public-webont-comments-request@w3.org]De la part de pat hayes
Envoye : jeudi 3 juillet 2003 00:23
A : bernard.vatant@mondeca.com
Cc : public-webont-comments@w3.org
Objet : Re: Could owl:sameAs reference non-OWL resources


  Dear Bernard


  This is in response to your comment/query

http://lists.w3.org/Archives/Public/public-webont-comments/2003Apr/0059.htm


  The short answer is 'yes'; that is, OWL can, in principle, interoperate
with 'non-OWL resources'. In fact there is not really a well-defined notion
of an "OWL resource" and hence not of a non-OWL resource: like RDF, OWL
simply uses URIrefs to denote things; and 'things' here can literally be
*any* thing at all; so the use of owl:sameAs to indicate that two things are
the same (more exactly, that two different URIrefs refer to the same thing)
is not restricted in any way: in particular, it is not restricted to "OWL
resources". (There is a restriction in OWL DL in that owl:sameAs cannot be
used there to refer to identity of OWL classes or properties: this is to
keep the OWL DL inference process within manageable complexity bounds. But
this restriction is private to OWL, as it were: it doesn't prevent OWL from
spreading its referential net widely across the universe of Web resources.)


  In the example you cite, the URIref "http//sports.org/US#SoccerTeam" is
supposed to indicate a class, probably the class of members of the team; so
this is indeed an externally defined resource. The OWL text asserts that
this is being regarded as an OWL class, which is merely OWL's way of saying
that it claims the right, as it were, to perform class reasoning about it,
eg think about the number of things in it, use it to help define other
classes, and so on.  But OWL is a general-purpose class reasoner, and it can
reason about virtually any class of virtually any kind of membership. It is
not restricted to a special category of "OWL classes" distinct from other
kinds of class or groupings. (OWL DL is somewhat restricted, eg it cannot
deal with classes containing other OWL classes, or classes of OWL
properties. Again, this is deliberate design decision which trades some
expressive power for a gain in processing speed in a reasoning engine.  But
any class that fits the OWL DL standards - which will be many of the cases
one meets with in practice - will be a suitable OWL-DL class.)


  I am not sufficiently familiar with XTM subject indicators to answer that
part of your question definitively, but certainly the OWL use of URIrefs
does not restrict them to those that retrieve a document which use any
particular format. Note however that your useage of the phrase 'the resource
at "http//sports.org/US#SoccerTeam" ' suggests that you think of this URI as
indicating the document 'at' the URL. OWL does *not* make his assumption: it
uses URIrefs simply as names, and assumes that they refer to something.
Exactly how these entities - the referents of the URIrefs - are defined is
not discussed by the OWL semantics. To this extent, OWL (and RDF and DAML)
are somewhat disconnected from the RFC 2396 picture of the use of URIrefs
within the REST architecture of an idealized Web. They do not deny this
architecture, of course, but (with a few exceptions, most notably the use of
owl:imports and the associated terminology) they largely ignore it. Thus,
there is nothing in the OWL spec that says how, if at all, one can
*determine* what class "http//sports.org/US#SoccerTeam" actually denotes:
the OWL stance on that question might be summed up as 'whatever class it is,
the following conclusions must be true of it, given what I know about
it...'; and OWL's task is simply to ensure that the process of drawing those
conclusions is error-free and as complete as possible. The same observation
applies to RDF and DAML; to this extent, these languages are similar in the
way they approach the meanings of URIrefs.


  Please let me know (with a cc: public-webont-comments@w3.org) whether this
reply adequately responds to your comment.


  Pat


--
---------------------------------------------------------------------
  IHMC       (850)434 8903 or (650)494 3973   home
  40 South Alcaniz St.       (850)202 4416   office
  Pensacola                       (850)202 4440   fax
  FL 32501                     (850)291 0667    cell
  phayes@ihmc.us       http://www.ihmc.us/users/phayes