- From: Peter F. Patel-Schneider <pfps@research.bell-labs.com>
- Date: Tue, 01 Oct 2002 08:32:40 -0400 (EDT)
- To: jjc@hplb.hpl.hp.com
- Cc: www-webont-wg@w3.org
From: "Jeremy Carroll" <jjc@hplb.hpl.hp.com> Subject: LANG, GUIDE, SEM: issue 5.9 Malformed DAML+OIL Restrictions Date: Mon, 30 Sep 2002 20:05:49 +0100 > SEM > http://www.coginst.uwf.edu/users/phayes/RDFS2OWL-F2F.html > (I realise this isn't the doc we were asked to read) > [[[ > I have changed the restriction conditions to IFF from Peter's document, as I > was worried that the conditions on restrictions may not be strong enough as > worded there. The issue is rather similar to the bug that Peter identified > in the RDFS subclass conditions: wording the semantics purely as only-if > conditions (if E is asserted, the interpretation must be thus...) does not > allow us to conclude that E is true, given some semantic conditions; the > 'iff' version seems better and clearly is stronger. The closure conditions > ensure that we will not be stymied by some class failing to be in the > universe, and the functionality of IRP secures a unique property for each > restriction class, so I can see no reason to have the weaker conditions. If > I'm not seeing some fatal problem, please let me know: it would be easy to > change it back. > ]]] > > Pat, > as I see it, your iff condition means that at least semantically any > owl:cardinality restriction would necessarily also have owl:minCardinality > and owl:maxCardinality arcs hanging off it. I am fairly agnostic on this, > but it does seem to relate to issue 5.9 > Jeremy Pat's new solution does not cause any problems that I can see with OWL restrictions. However, it does not work well in general. In particular, it would not work with qualified number restrictions, unless there was also some requirement that each qualified number restriction had a unique toClassQ, or some other way to outlaw strangely-constructed qualified number restrictions. If these sorts of conditions are required, why not just require that every restriction be ``syntactically well-formed'', i.e., exactly one onProperty and exactly one hasClass, toClass, hasValue, cardinality, minCardinality, or maxCardinality. peter
Received on Tuesday, 1 October 2002 08:34:05 UTC