Re: Issue-111 another angle and questions

On 28 May 2008, at 08:22, Alan Ruttenberg wrote:

> I alluded to there being another angle on the issue of signaling  
> user intent at our last meeting. Specifically I am thinking of the  
> case where there is necessity in addition to intent. So suppose you  
> are sending a message which is an OWL ontology, and which you need  
> to have interpreted correctly. Should there be a way  to be  
> indicate than unintended entailments should not to be made and/or   
> that the message can't be properly processed unless reasoning is  
> complete?  If so, does it make sense to have such a statement  
> anywhere but at the top level ontology O which is sent as the  
> message? If only at the top level, one thought is that such an  
> indication means: Ensure that the axiom closure is within the  
> syntactic bounds of indicated profile, and there are is no other  
> indication of necessary profile different from the one at top level  
> in any ontology in the imports closure.

I think, in general, necessity is a much more restricted case (as  
illustrated by your messaging example). For messaging (in particular)  
I would be inclined (as a semantic web services person) to think that  
the necessitating should come in the interface description (i.e., the  
WSDL). After all, you'd want to be able to coordinate that  
necessitation with the fault model (i.e., design a set of faults for  
various failures of inference). And that all sounds like it's rather  
out of our scope.

I think that's generalizable. Since necessity can only be determined  
*with regard to specific application*, the best we can due is a  
signal intent and required out of band agreements to determine the  
context. (That agreement can be as simple as your partners agree to  
respect your signaled intent!)

Cheers,
Bijan.

Received on Tuesday, 10 June 2008 17:48:48 UTC