W3C home > Mailing lists > Public > public-powderwg@w3.org > December 2007

URI groups

From: Stasinos Konstantopoulos <konstant@iit.demokritos.gr>
Date: Thu, 6 Dec 2007 16:26:38 +0200
To: public-powderwg@w3.org
Message-ID: <20071206142638.GH3561@iit.demokritos.gr>

Dear POWDER,

just my couple of cents on the URI groups issue. (This has nothing to do
with the equivalence vs subsumption thread.)

Andrea's proposal (current URI group document) looks like this:

<owl:Class rdf:ID="ResourceOnExampleDotOrg">
  <owl:equivalentClass>
    <owl:Restriction>
      <owl:onProperty rdf:resource="&wdr;includeHost" />
      <owl:hasValue>example.org</owl:hasValue>
    </owl:Restriction>
  </owl:equivalentClass>
</owl:Class>

(it doesn't really matter if it's going to be an implication or an
equivalence, the latter simply promises that there is no other way to
be in the ResourceOnExampleDotOrg class, like a DNS alias or
something.)

Now, I have already voiced my concerns about this approach on telcos
and emails, and I think I should re-iterate it, just for clarity's
sake. In a nutshell, it a well-known fact that once concrete domains
are added (strings, regexps, integers, and the like) it get really
hard to come up with a complete deductive operator and the RDF graphs
cannot be closed (ie, automatically generate the graph where all
implicit facts have been made explicit.)

In other words, there are important axioms defining the domain of URIs
that cannot be expressed in OWL, e.g. that a resource with value
"X.Y.Z" for its "includeHost" property also has value "Y.Z" and "Z"
for the same property. The domain of URIs cannot be closed under OWL,
because OWL is not expressive enough to imply property values given
other property values.

For this reason, the fact that ResourceOnExampleDotOrg has been
axiomatically defined in our example here isn't of much use, as it
pretends to specify all and every instances that are in the class,
where, in fact, it cannot.

My proposal is that URI Groups are defined in pretty much the same way
as reification: they're there, but not within the logic. Something
along the lines of:

  <owl:Class rdf:about="#FOSIchildSafe">
    <rdfs:subClassOf rdf:ID="CSassertion" rdf:resource="#ChildSafe"/>
  </owl:Class>

  <rdf:Statement rdf:about="#CSassertion">
    <foaf:maker rdf:resource="http://www.example.org/foaf.rdf#me"/>
  </rdf:Statement>

  <rdf:Statement rdf:about="#CSassertion">
    <wdr:includeHost "example.org"^^xsd:string/>
  </rdf:Statement>

This way we say nothing about subsumption or equivalenc and the such,
and an external, POWDER-specific tool will go over the DR and
explicitly add a resource to the FOSIchildSafe class, if appropriate.
OWL subsumption will then say that such a resurce is also ChildSafe.

BTW, with OWL semantics out of the way, one can possibly even go back
and throw excetions and excludeHost statements back into the game.

Hope this helps a bit,
stasinos
Received on Thursday, 6 December 2007 14:27:01 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:42:12 GMT