[Fwd: Re: here's a draft charter strawman slideset, wrt W3C Rules WG]

-- 
Francois Bry
http://rewerse.net  scientific coordinator
http://www.pms.ifi.lmu.de  head of unit 

Forwarded message 1

  • From: Jos de Bruijn <jos.debruijn@deri.org>
  • Date: Wed, 29 Jun 2005 16:10:44 +0200
  • Subject: Re: here's a draft charter strawman slideset, wrt W3C Rules WG
  • To: Francois Bry <bry@ifi.lmu.de>
  • Message-ID: <42C2ABE4.6010603@deri.org>
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Dear Francois,

I assume you meant to include the mailing list in CC in the previous
email (you apparently forgot to do so), unless you wanted to respond
in private.
If indeed you intended to address the mailing list, feel free to
forward my reply to your replies to the mailing list :)

Francois Bry wrote:

> Dear Jos,
>
> Thank for your answer to my comments! May I answer to them?
>
>>> 2. I do not clearly recognise the central issue "What rules
>>> languiages what for". "Rules" means many different things that , I
>>> beleive, are better kept distinct:
>>>
>>> 2.1 views, or predefined queries (rfegardless of whether they are
>>> evaluated by forward of bac kwartd chaining, I fully agree with the
>>> proposal in this respect).
>>>
>>> 2.2 Integrity constraint-like rules, i.e. FOL-like or Description
>>> Logic-like, eg OWL, formuylas evaluated against a repository of
>>> factual data (eg database).
>>>
>>> 2.3. Ontology-like rules, i.e. FOL- or Description LOic-likre, eg
>>> OWL formulas considered independently of any repository of factual
>>> data.
>>>
>>
>>
>> I would say 2.3 incorporates both 2.1 and 2.2. OWL incorporates both
>> derivation rules (e.g. subclass axioms) and integrity rules (e..g.
>> disjointness axioms).
>>
>>
> Yes and no. In terms of expressive power: Yes. In terms of
> processing: No, for two reasons:
>
> 1. No OWL reasoner will properly (i.e. efficiently) evaluate
> integrity constraints against several WEb sites. OWL reasoner just
> do not have the capability to do it.
>
> 2. The negation issue is part of the answer. Consider an ontology
> defining when a student is allowed to register for an exam (a German
> examples because at Germasn universities there are complex
> regulations for this). The specification will for sure include
> negative (sub)formulas. If the regulation is considered for itself,
> eg for checking whether it is consistent or makes it possible for a
> student to graduate within a given number of terms, negation will
> have to be interpretaed as classical logic negation, i.e. excluded
> middle and/or refutation make sense. If the regulation is used for
> chercking against a database of student who enforce/ does not
> enforce the regulation, the very same negated subformula will have
> to be understiood using (possibly scoped) nonomonotonic (or default)
> negation (of negation as failure, all these denomination are
> equivalent at this level).

I actually agree with you on both points, as I have argued in [1].

>
>>> 2.1 and 2.2 (since integrity constraints are perfectly realized as
>>> special views) pertain to Query Languages. Therefore, any activity
>>> on a W3C rule language for the SW should be tightly joined with
>>> activities on SW Query language(s).
>>>
>>
>>
>> Completely agree. Rules and queries have always been closely related.
>> Typically, a query is a rule without a head.
>>
>>
> I believe, this should clearly show up in a charter.

For the charter, we should be careful how we formulate the
relationship with queries, since there are already other working
groups tackling the issue, but the relationship should indeed be
described.

>
>> There has already been a lot of discussion on how to integrate OWL and
>> rules. The two leading (well-known) ways to integrate them are:
>> - - use of a common subset; disadvantage is that many (especially DL
>> advocates) believe that this subset is too small for any useful
>> ontology
>> - - exchange of consequences between the OWL KB and the rule base
>> (e.g.
>> dl-programs by Eiter et al.)
>>
>>
> I know, but I feel uneasy about these proposals, because they are
> nmot clearly defining what for "rules" should be used for.

I'm not sure I understand. Do you mean the kind of rules which are used?
In the common subset approach we are talking about a restriction of
Datalog.
In Eiter's approach we are talking about full LP with either Stable
Model Semantics or Well-Founded Semantics (dl-programs were defined
for both semantics).

>
>> As I already said, OWL contains integrity constraints.
>
> Again: as formulas, but not the processors for their evaluation
> against factual data.
>
>> However, they
>> don't necessarily behave as you might expect, since they are not used
>> to check the integrity of the data, but rather to constrain the number
>> of first-order models.
>>
>>
> Here we are. But I am convinced thnat both usages are needed on the
> Semantic Web.
>
>> It would, IMHO, be interested to see how we could relate OWL
>> (open-world) constraints, which are very helpful when reasoning over
>> the schema, and closed-world (and UNA) integrity constraints which can
>> be used to check a set of data.
>
> It seems to me that the above could be seen as a beginning of a
> sketch answering the question?

well, we are not talking about a general solution to relate DL and
rules. But as I argued in an earlier email, it might be worth
considering this for OWL2.

>
>> However, I don't think this should be
>> the focus of the WG.
>>
>>
> Well, the focus should be derived from the kinds of rules that
> are/will be needed in practice. I beleive: db-like integrity
> constraints, and ontology-like froumals are both needed (that
> furrthewrmore could syntactically be identical, in some cases).

I agree that both kinds of constraints are probably needed. IMO, the
rules language should definitely allow for integrity constraints, but
interoperation with OWL is still a hairy issue here. It would be nice
if OWL2 moved a bit in this direction; this would facilitate
interoperability.

>
>> IMO, active/reactive rules should not (yet) be the focus of the WG. I
>> think we should first tackle the problem of derivation and integrity
>> rules.
>>
>>
> That would surely help in having a clear focus. But a questionj is
> then whether active/reactive rules will not be needed at some later
> point? And many workshop participants and news list participant
> explicitely refer to active/reactive rules...

Yes, they might be needed at a later point, but we should not try to
solve everything in one go.


Best, Jos

>
> Regards,


[1] Jos de Bruijn, Axel Polleres, Rubén Lara, and Dieter Fensel: OWL
DL vs. OWL Flight: Conceptual Modeling and Reasoning for the Semantic
Web, /Proceedings of the 14th International World Wide Web Conference
(WWW2005)/, Chiba, Japan, 2005.

- --
Jos de Bruijn, http://www.uibk.ac.at/~c703239/
+43 512 507 6475         jos.debruijn@deri.org

DERI                      http://www.deri.org/
- ----------------------------------------------
Clothes make the man.  Naked people have
little or no influence on society.
        -- Mark Twain
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFCwqvj4lqeiwiiHN4RAlyXAJ4jEnBmwZUi3r1El9PMwWsWMsKIAACcCsuj
1h8mlpAHRWMDSABSkylIWHA=
=AMMj
-----END PGP SIGNATURE-----

Received on Wednesday, 29 June 2005 14:32:27 UTC