- From: Graham Klyne <GK@ninebynine.org>
- Date: Wed, 07 Mar 2001 12:13:35 +0000
- To: "Tim Berners-Lee" <timbl@w3.org>
- Cc: <www-rdf-interest@w3.org>
I think this is an excellent example of why we should not rush to put full first order logic into the core of RDF, but rather to build it, where needed, in a framework constructed on RDF. I have a sense that for some applications, like this, there is much that can be achieved without full FOL. And there may be other ways of dealing with things normally handled by negation. For this reason, I agree with your assertion (made elsewhere) that A rdfs:subClassOf B should allow A and B to be the same. A small terminological point for consideration: in my own mind I have taken to distinguishing between 'infers' and 'implies', such that 'infers' is the result of applying rule forward chaining in a world that may not contain negation, and implies (A=>B) is the logical assertion ((NOT A) OR B). #g At 11:25 PM 3/6/01 -0500, Tim Berners-Lee wrote: ><~/calendar.rdf> log:impliesByFPC { :i a cal:OverCommittedPerson } . > >means that there is a forst order proof of the conclusion from the evidence >in the document. > >Now you can be specific about negation > ></etc/passwd.rdf> log:doesntSay { :<#timbl> unix:inGroup <#www> } . > >we can state a default in a well defined way: > >{ </etc/passwd.rdf> log:doesn'tSay { :x unix:inGroup :y } } log:implies > :x unix:inGroup <#others> } . ------------ Graham Klyne GK@NineByNine.org
Received on Wednesday, 7 March 2001 07:36:38 UTC