- From: Jacek Kopecky <jacek.kopecky@deri.org>
- Date: Thu, 22 Feb 2007 21:52:55 +0100
- To: Richard Cyganiak <richard@cyganiak.de>
- Cc: Hamish Harvey <hamish@hamishharvey.com>, semantic-web@w3.org
Richard, let me try a take: On Thu, 2007-02-22 at 21:01 +0100, Richard Cyganiak wrote: > On 22 Feb 2007, at 20:17, Hamish Harvey wrote: > >> Oh! I have to admit that I did not think that far. By "relational > >> schemes" I just meant collections of table definitions by their > >> attributes, together with some basic inter-table relationships like > >> foreignkeys. Now, with such general integrity constraints in mind, > >> the > >> whole discussion might perhaps get quite another direction. Hm... > > > > But those integrity constraints _enforce_ a certain semantic > > consistencies. They do not _represent_ semantic relationships as such, > > do they? > > What's the difference? Enforcement means, if you successfully put in some data, it's there. Representing semantic relationships here (and perhaps the word "inference", but I'm not sure) would mean that if you put in some data, the result contains more data than just what was there already and the new, there are also new inferences. Typical difference between typing constraints and OWL's range: Person hasParent Person I want to add "Joe hasParent Jane", but I can't do that with typing constraints, I'd have to say "Joe isa Person. Jane isa Person. Joe hasParent Jane." In OWL, I can say "Joe hasParent Jane" and OWL will tell me "Joe isa Person. Jane isa Person." Some applications require type checking, some will benefit from inference, both should be supported. For instance, WSMO (wsmo.org) has both. Best regards, Jacek
Received on Thursday, 22 February 2007 20:53:22 UTC