Re: [closed] pfps-05 RDFS Closure Rules

>From: pat hayes <phayes@ai.uwf.edu>
>Subject: [closed] pfps-05 RDFS Closure Rules
>Date: Mon, 21 Apr 2003 18:27:55 -0500
>
>>  Peter,
>>
>>  Re. your comment
>>
>  > http://lists.w3.org/Archives/Public/www-rdf-comments/2003JanMar/0090.html
>>
>>  The editor has accepted your comment and the closure rules have been
>>  modified to cover this (and other ) cases, by incorporating
>>  existential generalization as an explicit closure rule.
>>
>>  Please reply to this email, copying www-rdf-comments@w3.org indicating
>>  whether this decision is acceptable.
>>
>>  Pat
>
>This decision is not acceptable.
>
>It is not acceptable for structural reasons. 
>
>First, no indication of the changes that have been made are accessible from
>this message.

Right, I short-circuited that slightly, knowing that you had access 
the current draft with the changes in it. But I hereby slap my wrist.

>Second, the comment was not just about existential closures,
>but instead was about whether the closure procedure for RDFS are complete and
>whether the RDFS entailment lemma is true.  The response above does not
>address this portion of my comment.

OK, I will reply later with a more detailed official response. 
Apparently my reply was not following official WG protocols in any 
case, so it needs to be re-sent. I never was good at following 
irrational procedures.

>My investigations of the current editor's draft of RDF Semantics (21a) also
>indicate that there remain in this draft problems related to my comment.
>
>First, the semantic conditions for RDFS uses IC(rdfs:Container), which is
>ill-defined.

Indeed, a slip I had not noticed, thanks. Fixed.

>  If this is changed to the meaningful ICEXT(rdfs:Container),

ICEXT(I(rdfs:Container))

>the condition is not appropriate.  Even if this is changed from a range to
>a domain,

another typo, thanks: I had the range and domain backwards. Fixed.

>I believe that it is still not appropriate because I believe that
>the intent in RDFS is that container membership properties can be applied
>to non-containers.

That is not my understanding. I will refer this to the WG for 
confirmation one way or the other. Can you document this intent?

>
>Second, the RDFS closure procedure says to add the triples from a table and
>some other triples.  However, these other triples are already in the table.

Right, another editorial slip left over from an earlier draft. Fixed.

>
>Third, the closure rules are unable to infer any rdfs:domain or rdfs:range
>triples.  It thus appears to me that the RDFS closure procedure is still
>incomplete as it will not include triples of the form
>
>	rdf:_n rdfs:range xx .
>
>(for whatever xx is deemed to be appropriate).

I do not think this is an error in the rules. The domains and ranges 
for all the properties in the whole RDFS vocabulary are now included 
in the axiomatic triples table explicitly, and no other domains and 
range statements can be concluded using the RDFS semantic conditions 
(except possibly by inference paths involving subPropertyOf, which 
are covered by the rules).

However, I have re-phrased the text immediately after the statement 
of the lemma ('captured'/justify) and added a sentence to emphasize 
that the entailment lemma as stated would not generalize to the case 
where the strengthened domain and range conditions apply and the 
modified rule set with rules 2a, 3a 4a' and 4b' are used.

>I note also that there are many changes to the RDF Semantics.  These
>changes may have uncovered previously unnoticed issues and may themselves
>have issues.

If you have any issues, please raise them. The changes are mostly 
rearrangements rather than substantial changes - some conditions on 
the RDF vocabulary have been moved back from RDFS to RDF - and of 
course there may be typos or omissions arising from those editorial 
changes to the document.

Pat
-- 
---------------------------------------------------------------------
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@ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
s.pam@ai.uwf.edu   for spam

Received on Tuesday, 22 April 2003 12:49:25 UTC