RE: Objective 4.6 -- additional semantic information

I think we clearly need to find some common ground, because I object to
4.6a and am fine with 4.6

I understand both of your objections, but to be honest they are the main
*reasons* I think this objective is desirable. I think it's important to
point out that even though there are ways to encode knowledge about RDF
graphs that have not yet been standardized, we strive to produce a query
language in which *any* extra knowledge which describes RDF graphs can
play a role. SWRL is the best example I know of; if adding one or two
others (remember, these are just examples, not an exhaustive list)
alleviates your concern that we're explicitly endorsing SWRL, then maybe
that's a solution.

I'm also interested in your "confusing rule processing with other kinds
of inferencing" statement. So is rule processing a kind of inferencing?
How do you see it as fundamentally different from other kinds of
inferencing? I'm not sure whether we're coming back to the "inferencing
is nothing more than adding triples" argument, but if so it is quite
clear that even OWL takes you beyond this: there is no canonical set of
triples which represents all consequences of an OWL ontology within RDF
using RDF semantics.

I very much intended this objective to be vague and non-required of
implementations, keeping in mind that most "inferencing" systems are not
based on sound and complete reasoning algorithms. Further, the actual
details of such reasoning (and even the format of knowledge encoded
beyond bare RDF triples) are explicitly beyond the scope of this group.
Any SHOULD would be burdening simple RDF processors with the requirement
to process all the other semantic languages in the universe. The
objective is that systems with more robust reasoning capabilities
available should be able to use them to assist in answering queries
about the RDF graph(s) they describe. If you feel that we need to be
more specific in terms of soundness versus completeness in order to
enforce monotonicity of query results, then we should talk about an
appropriate wording. My opinion is that it's more detail than anything
else in this document addresses, so it's not really necessary at this
stage.

> -----Original Message-----
> From: public-rdf-dawg-request@w3.org 
> [mailto:public-rdf-dawg-request@w3.org] On Behalf Of Jim Hendler
> Sent: Tuesday, June 08, 2004 1:17 PM
> To: public-rdf-dawg@w3.org
> Subject: Objective 4.6 -- additional semantic information
> 
> I have now reached the end of the UC&R document and would 
> like to register that I would object (in the sense that if 
> this went to rec I would vote against it as an AC member if 
> it included these words) to 4.6 as written:
> 
> It should be possible for knowledge encoded in other semantic 
> languages-for example: RDFS, OWL, and SWRL-to affect the 
> results of queries executed against RDF graphs.
> 
> while I could live with 4.6a
> 
> It should be possible for a query to indicate that the 
> answers should take into account knowledge encoded in RDF 
> semantic extensions such as RDFS, OWL, etc.
> 
> This is for two reasons:
>  1 - mention of SWRL is problematic -- RDFS and OWL are 
> Recommendations that have gone through the full W3C process 
> at great cost to participants and to the W3C.  SWRL is just a 
> note w/no WG status.   While I like SWRL, and my research 
> group uses it, there's a bunch of other systems out there at 
> the same level of informality, and it isn't fair for us to 
> mention one and not the others (some of which are notes or 
> referenced in notes).    Also, adding SWRL confuses rule 
> processing with other kinds of inferencing, and it is not 
> clear to me this is an objective of this WG (in fact, I would 
> point out that it is probably out of scope given section 2.2 
> of the  charter [1])
>  2 - the "affect the results of queries" seems to me to be 
> not only imprecise, but misleading -- it doesn't say 
> "enhance" or "improve" so it could simply mean that that 
> stuff can arbitrarily be taken into account in some way -- 
> FWIW, I could achieve what is stated in 4.6  by saying 
> whenever RDFS or OWL is used the query fails -- which means 
> the knowledge has affected the queries.   Seriously, I think 
> the second writing, which offers that the SHOULD take into 
> account the knowledge, is much clearer as to what the 
> objective actually is
> 
> [1]  http://www.w3.org/2003/12/swa/dawg-charter#rules
> 
> -- 
> Professor James Hendler                   
> http://www.cs.umd.edu/users/hendler 
> 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      240-277-3388 (Cell)
>    
>

Received on Tuesday, 8 June 2004 17:27:07 UTC