Re: How core is Core?

Michael Kifer wrote:
>> An alternative is to say that Core is not core it is just a starting 
>> point for dialect design and concrete dialects will pick and choose from 
>> Core as appropriate. That's what I understood to be MichaelK's position.
> 
> Yes. I think it would be hard, if not impossible, to have a useful core
> that everybody would be required to implement. With this requirement, we
> would either have to come up with something uninteresting or with something
> that only a few will implement.

On the telecon today Christian tried to approach the question of whether 
a shared Core is useful by asking for cases where you might want to 
translate a PR ruleset to an LP setting.

I would like to approach the question differently.

I claim that an "interesting" range of uses of rules on the semantic web 
can be covered by function-free horn rules.

My evidence for this is:
   (a) both N3 rules and JenaRules seem to get significant use [*]
   (b) a number of SWRL users (though by no means all) who turn up on 
jena-dev seem to only be looking to implement the "rule bit" of SWRL 
without the OWL entailments
   (c) specific uses cases such as the vocabulary translation use case 
(UC8) are covered by this

This subset of Core is implementable in both production rule and LP 
settings.

Now of course RIF's job is not to define a rule language for the 
semantic web so this is by no means a clinching argument. I'm simply 
pointing out that a RIF subset so boringly simple that many languages 
could implement it might still have *some* applications.

This is not advocacy. I'm happy to continue in the current direction and 
have Core not be truly core.

Dave

[*] OK the argument is a little weak because a non-trivial number of N3 
rules use things like log:semantics and of JenaRules use negation over 
base data (noValue).

The ability to partition DataSets (as discussed separately), query those 
partitions and ultimately perform closed-world negation over those 
partitions would address many of those uses. That would still be 
implementable in both LP and PR settings.

-- 
Hewlett-Packard Limited
Registered Office: Cain Road, Bracknell, Berks RG12 1HN
Registered No: 690597 England

Received on Tuesday, 3 July 2007 21:25:07 UTC