Re: [RIF] [UCR]: What is the RIF (revisited) --> disjunctive conclusions

> 
> Ed Barkmeyer wrote
> 
> >The point is that a "business rule" taken out of an interpretation context can 
> >be highly ambiguous.  
> >
> Right. But this "out-of-context" use is an extremely important issue. 
> THe Web has been successfull to a largfge extent because it is based on 
> making an "out-of-context" use possible. My strong conviction is that 
> RIF should make such a use easy.
> 
> Consider again my example: "every student with major in Computer Science 
> must have a minor in Mathematics or in Physics".

I believe that when people talk about disjunctive heads, they mean
deductive rules. Integrity constraint like the above are represented as
queries in LP and DBs. It is a query with an implication in the body, which
has a disjunction in the head (of that embedded implication). Lloyd-Topor
transforms it into constraints with no disjunction or implication.
So, this is not the kind of rules that Dieter meant to exclude, I believe.

I think we have a consensus that we should not tackle disjunctions in the
heads of *deductive* rules in Phase 1. Perhaps not even in Phase 2.
However, we *should* design RIF in such a way that such extensions would
be possible.

Since we already discussed that RIF rulesets could be tagged with semantics
to let the recipient understand the intended meaning, I don't see
significant obstacles to allowing disjunctions in the heads of deductive
rules when these are tagged with classical or stable-model semantics. The
recipient engine can reject such rules, if it doesn't have an engine to
process them.

This mechanism doesn't seem to be too controversial to me and I don't quite
understand what all the fight is about.


	--michael  

> The rule is prefectly clear to everyone (assuming an understanding of 
> the words used). The rule can be used as an integrity ocnstraints (ie 
> for checking if every strudent enforce it), the rule can be used 
> combined with other regulations (=TBoxes) for deriving interewsting 
> conclusions without considering student data (=Aboxes).
> 
> RIF should not exclude any use.
> 
> -- 
> Francois
> 
> 
>

Received on Thursday, 9 February 2006 13:48:08 UTC