- From: Jeremy Carroll <jjc@hplb.hpl.hp.com>
- Date: Tue, 25 Mar 2003 22:50:14 +0000
- To: "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>
- CC: www-webont-wg@w3.org
Jeremy: >> A more attractive option, requiring more detailed >>work, would be to inspect each use of the disjoint bnodes assumption in the >>proof, and make necessary changes. >> Peter: > I was planning on doing something along this latter line. Given the other > demands on my time I can't promise to sufficiently attend to this until > mid-April, as I stated in a recent telecom. > I have just been looking at it - it looks challenging. [Rest of message rambles ...] The proof often combine bnodes mappings A1, A2, .. An into a single bnode mapping relying on the disjointness of the bnodes. In the non-disjoint case these could be combined with a rule A*(n) = Ai(n) for the first Ai with n in its domain we intuit that since CEXT(Ai(n)) = CEXT(Aj(n)) in the interesting cases, and the semantics of descriptions really only depends on their class extension, that everything will be OK - but ... the lists look hairy for example. The individual list nodes join specific domain elements so this naive A* will not work. Jeremy
Received on Tuesday, 25 March 2003 17:51:44 UTC