Re: Error in current WG test cases?

>pat hayes wrote:
>
>>>In a message on 25th July [1] I noted that the tests:
>>>   rdfms-seq-representation/Manifest.rdf#test002
>>>   rdfms-seq-representation/Manifest.rdf#test004
>>>appeared to be incorrect.
>>
>>
>>No, they are correct.
>
>The RDFCore message thread [2] responding to my previous reporting 
>of this seemed to indicate otherwise, I guess I misunderstood or 
>something has changed.
>
>>The stuff mentioned in the conclusion doesnt have to be mentioned 
>>in the assumption.  However, it is fine for a FORWARD rule-based 
>>reasoner not to generate all valid conclusions from an empty graph, 
>>just as it would make sense not to generate everything from a 
>>contradiction. But think of a query-responding engine that is given 
>>the conclusion as a query: it could just say YES without even 
>>bothering to check into its KB.
>
>We're not using a (purely) forward rule-based reasoner, we are 
>running the tests via a query interface, and would have no problem 
>implementing it, if that is the correct thing to do.
>
>My reading of the LC2 working drafts (not the editor's draft) is 
>that a query to an empty model of the form:
>
>   * rdf:type rdfs:ContainerMembershipProperty
>
>should return empty, due to the restriction of the rdfs axioms to 
>the vocabulary {V U crdfV U rdfsV}. Is that right (again in the LC 
>docs not the editor's draft)?

I'm not sure what semantics you intended for querying, but if the 
question is posed in terms of entailment, and if you supply an actual 
CMP name and ask if it is a CMP, then the answer should be 'yes' even 
from an empty graph, since (in LC) entailment(of E by S) is defined 
relative to interpretations of the combined vocabularies of E and S.

>I guess this would be framed as a negative entailment test of the form:
>
>    "does the empty graph entail the graph:
>    _:a rdf:type rdfs:ContainerMembershipProperty ."
>

OK, in that form the answer is "yes" both in the LC version and the 
modified version. It does entail it.

>If this understanding is right then that is why I felt the 
>vocabulary in the conclusion does have to be mentioned in the 
>assumptions. It would be rather inconvenient if a query of this form 
>returned empty but a query for a specific rdf:_1 property returns a 
>match.

Well, the MT doesnt specify a semantics for a query language, so I 
don't quite know how to deal with this. BUt take as an analogy simple 
propositional calculus, where any tautology is entailed by the empty 
set of assertions. Given an empty set, you wouldn't want an engine to 
generate all tautologies: but given a a tautological query, you might 
expect it to say 'yes'. What about given part of a tautological 
query? Suppose you ask it

* or not P ?

would you expect it to answer 'P' ? I wouldn't, myself: but in any 
case, the semantics has no way to answer that question, since you 
havn't give a semantics for the wild-card symbol.

Being more helpful, I think that a perfectly reasonable way of 
interpreting the wild-card is that it should match any vocabulary 
items actually present, so that it would make sense for such a query 
to fail on the empty graph but succeed when an actual name is 
supplied.  That would be normal for a syntactic wildcard. It means 
that a wildcard is not the same as a genuine universally quantified 
variable, but then you probably don't want it to be.

>I guess all this is now moot. I see from [3] that the semantics 
>document has changed to drop the crdfV restriction

Yes....

>so the above wildcard query should return an infinite set of 
>matches. Is that correct?

I really have no idea what you think the appropriate behavior for a 
wildcard query should be; I don't think we have ever discussed this 
in the WG (??to my recollection, in any case). The test-case 
entailments havn't changed, however.

Pat

>
>This seems like quite a big post-LC2 change.
>
>Dave
>
>[2] http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2003Jul/0324.html
>[3] http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2003Nov/0041.html


-- 
---------------------------------------------------------------------
IHMC	(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.	(850)202 4416   office
Pensacola			(850)202 4440   fax
FL 32501			(850)291 0667    cell
phayes@ihmc.us       http://www.ihmc.us/users/phayes

Received on Wednesday, 12 November 2003 11:51:03 UTC