RE: ISSUE 5.2 Language Compliance Levels - proposed clarification

On January 29, Jim Hendler writes:
> At 15:30 +0000 1/29/03, Ian Horrocks wrote:
> >On January 27, Jeremy Carroll writes:
> >>
> >>  Thanks Ian for this pointer - it does seem highly relevant to the content of
> >>  my proposal.
> >>
> >>  > 4. If you really did succeed in eliminating the ability to express
> >>  > "complete" classes in OWL Lite, you would make it useless in a wide
> >>  > range of important applications (e.g., see [3]).
> >>
> >>  > [3] http://lists.w3.org/Archives/Public/www-webont-wg/2002Dec/0088.html
> >>  >
> >>
> >>  I repeat one part of that:
> >>
> >>  [[
> >>  We have also done a lot of work recently on a publish and subscribe
> >>  system using DAML+OIL/OWL. This is similar to the above service
> >>  discovery application in that subscribers describe the kinds of
> >>  "publication" (e.g., messages) they are interested in, and messages
> >>  are routed to subscribers according to their descriptions.
> >>  ]]
> >>
> >>  If I have understood correctly, without the complete class descriptions the
> >>  subscriptions could not be made. For instances if I want messages both about
> >>  HP and the SemanticWeb, I can say that the messages I want are subClassOf
> >>  both of these, but without the complete part of the class description any
> >>  particular message that has been categorized as in both, may fail to be in
> >>  my subset of the intersection.
> >
> >Correct.
> >
> >>  Personally, I think we could decide that publish and subscribe type
> >>  applications need to use OWL DL; but I emphasis - I want to concur with the
> >  > majority here.
> >
> >My original message ([3]) was in support of my strong disagreement
> >with this statement. Why do you want to impose the *significant*
> >additional overhead of OWL DL reasoning on a very wide range of
> >applications?
> >
> >Ian
> 
> Ian - could you actually expand on this a bit -- in particular, the 
> current Owl Lite and the Current Owl DL are in the same complexity 
> class, except for OneOf -- the additional implementational burden 
> from Lite to DL seems fairly small if one doesn't include these 
> features based on our reading about tableaux (and not including 
> datatypes which are as yet unresolved in our design) - are we wrong? 

Er, this seems a bit like saying that going from propositional logic
to predicate calculus is easy if you don't support quantifiers. Having
said that, you are correct in saying that deciding
satisfiability/entailment in Owl DL without OneOf has the same
complexity as Owl Lite.

> We're really trying to figure out if writing an OWL Lite reasoner is 
> really easier than writing a SHIQ reasoner - and from the Handbook of 
> DL it looks like it isn't that much different.
>   (re-reading the above it looks liek I'm advocating something else -- 
> let me make it clear the above is meant as a real question -- can you 
> elaborate a bit on the additional complexity of moving from Lite 
> Reasoning to DL reasoning - as it effects a lot of how I write our 
> implementation report)
>   -JH

Well, Lite doesn't have negation and disjunction, so it must be
much easier, right?

Seriously, as we already established, it is in the same complexity
class as SHIQ, and one could use the same basic technique. In
practice, the lack of (qualified) number restrictions would make OWL
Lite significantly easier to implement (it is more like SHIF).

Ian

> -- 
> Professor James Hendler				  hendler@cs.umd.edu
> Director, Semantic Web and Agent Technologies	  301-405-2696
> Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
> Univ of Maryland, College Park, MD 20742	  240-731-3822 (Cell)
> http://www.cs.umd.edu/users/hendler

Received on Thursday, 30 January 2003 11:52:49 UTC