Re: Web Rule Language - WRL vs SWRL

On Jun 29, 2005, at 1:58 PM, Ian Horrocks wrote:

> On 28 Jun 2005, at 20:17, Michael Kifer wrote:
[snip]
>> N3 is essentially a different syntax for F-logic and its extensions 
>> (but
>> N3's semantics is defined by use cases ;-).

Test cases, actually. At least, that's what some say.

>>  As far as I can tell, with each
>> new presentation that I hear N3 is moving in the direction of LP.
>
> I think that we should stick to discussing how things actually *are* 
> rather than directions in which you hope/believe they might be moving. 
> Surely N3 is *actually* a different syntax for RDF, plus some 
> rule-like extensions.
[snip]

I took Michael's F-Logic remark to be focused on the HiLog/Common 
Logic/collection theory aspects of it (i.e., first order 
semantics/second order syntax) and the fact that N3 proponents 
sometimes claim to be Horn oriented. Of course, with Bnodes in the head 
and the possibility (though it's unclear) of function terms, the exact 
expressiveness is hard to pin down. (I hope that the PAW project will 
nail it down so that Tim's intent is made clear, perhaps even to 
himself :))

Of course, in many lights, N3 doesn't look like SWRL in that it 
certainly doesn't take OWL seriously. There is some evidence of intent 
to rule out disjunction, nonbnode existentials, and some other things. 
There is also evidence that the interpretation of cardinality might go 
databasey if pressed. Hard to say.

Be all this as it may :)

I'll add that N3's attempt (perhaps the way OWL Lite attempts) to avoid 
certain computational complexity inducing features is closer to what 
Michael might advocate, to wit, some form of Datalog with variously 
restricted default negation. I.e., I don't see michael advocating 
Disjunctive Datalog with answer set semantics any time soon.

The third corner in all this would be those, perhaps someone like Pat 
Hayes, who might advocate that the next effort should define as 
expressive a language as seems reasonable for the next, oh, 5 years. 
Full FOL as a minimum. And then subset. (LBase, anyone?) This would be 
distinct from the agenty/TRIPLE models/query interface/black box model 
that Michael mentions, and rather in opposition. I would imagine that 
someone like Pat might argue (heh, Pat! jump in anytime :)) that the 
latter approach is best conceived as an implementation of the former 
approach, where the former approach is seens as the specification. This 
is not unlike his approach to nonmonotonicity referenced earlier in 
this thread.

So, that's where I'm headed. I would like to see a task force, or 
working group, or *something* deal with the integration framework. I 
mean, I'll keep puttering on it natch, but I think anyone proposing a 
standardization effort owes something reasonably sketched or fleshed 
out. The old Layering Story has been bankrupted in several different 
ways.

Cheers,
Bijan Parsia.

Received on Wednesday, 29 June 2005 18:33:38 UTC