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

SEM: 5.5 List semantics

From: Jeremy Carroll <jjc@hplb.hpl.hp.com>
Date: Mon, 28 Oct 2002 10:18:30 +0100
To: <www-webont-wg@w3.org>
Message-ID: <BHEGLCKMOHGLGNOKPGHDIEICCAAA.jjc@hpl.hp.com>


RDFCore discussed this on Friday:
http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Oct/0386.html
[[
discussing semantics of RDF Collection

Jeremy  feels strong about this e.g. lack of equality,
        also possibility of contradiction
DanB    if we add notion of a functional property for lists,
        i'd want it across the board... (hmm scope creep...)

ACTION: DanC to produce a testcase for a nonentailment
ACTION: DanC to explain why it's ok not to put functional semantics
        for rdf:first/rest in our spec.
]]

Pat seemed to bow to the pressure of strong opinions from me and Dan.


The result is that the RDF Model Theory will say little or nothing on the
semantics of rdf:List. Thus the semantics required by OWL will need to be
covered in the large OWL MT.

There have been some attempts to show how in part that can be achieved.
These rely on using a class owl:List that is a subClassOf rdf:List, and
using  range constraints to make it all hang together.
(Note: the RDF/XML syntax will only deliver rdf:List).

Problem statement from Dave Beckett:
http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Oct/0382.html
[[[
So, assume there is an owl:List class:
  owl:List rdfs:subClassOf rdf:List.
which OWL uses for some purpose, not important here.  That's fine.

However, the convienence RDF/XML syntax we have for these collections
(which, as we are discussing, may have no particular semantics in RDF)
is as follows:

  <ex:prop1 rdf:parseType="Collection">
    <rdf:Description rdf:about="uri1"/>
    <rdf:Description rdf:about="uri2"/>
    <rdf:Description rdf:about="uri3"/>
  </ex:prop1>

which gives a big pile of triples:

  _:node ex:prop1 _:genid1 .
  _:id1 rdf:type rdf:List .
  _:id1 rdf:first <uri1> .
  _:id2 rdf:type rdf:List .
  _:id2 rdf:first <uri2> .
  _:id1 rdf:rest _:id2 .
  _:id3 rdf:type rdf:List .
  _:id3 rdf:first <uri3> .
  _:id2 rdf:rest _:id3 .
  _:id3 rdf:rest rdf:nil .

but the triples for the nodes in the middle of it are of rdf:type rdf:List.

There is no way for those to be made into owl:List using this form of
abbreviation, hence no way for OWL to have a convienent notation for
what might be called OWL lists.
]]]

I suggest the next steps could be:
- get DanC or myself to propose a detailed account of how rdf:List's can be
turned into owl:List's using range constraints.
- ask Pat to propose a suggest semantics for owl:List that ensures that the
non-FOL issues are addressed (I don't pretend to understand these, something
to do with either living with, or prohbiting non-standard models with
infinite owl:Lists or circular owl:Lists etc).

Jeremy
Received on Monday, 28 October 2002 04:18:31 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:53 GMT