Re: [proposed close: pfps-05]

The last-call version of RDF Semantics contains many significant errors.  I
have pointed out many of these.  I even pointed out some of them in the
last-call version before RDF Core went to last call.  Nevertheless, the RDF
Core WG decided to go to last call with these comments unaddressed.

One of these signficant errors was that the RDFS closure rules were
incomplete.  I have pointed out problems with the closure rules in various
versions RDF Semantics, starting on 26 September 2001
(http://lists.w3.org/Archives/Public/www-rdf-logic/2001Sep/0031.html).  I
reported problems with the closure rules in the last-call candidate on 27
December 2002
(http://lists.w3.org/Archives/Public/www-rdf-comments/2002OctDec/0297.html).

On 24 January 2003, I made a formal last-call comment concerning the
completeness of the RDFS closure rules
(http://lists.w3.org/Archives/Public/www-rdf-comments/2003JanMar/0090.html),
which states

>The closure rules for RDFS are incomplete.  For example,
>    ex:foo ex:prop "a" .
>RDFS-entails
>    ex:foo ex:prop _:x .
>    _:x rdf:type rdfs:Resource .
>However, this does not come out of the RDFS closure rules.
>This means that the RDFS entailment lemma is false.

Since then many changes have been made to the RDF model theory that have
affected this issue.  In each version of RDF Semantics up until at least
the 15 July 2003 version the RDFS closure (or entailment) rules have been
incomplete and the RDFS entailment lemma has been false.  I have
repeatedly pointed out these problems.

In the 31 July 2003 version of RDF Semantics, the RDFS entailment rules are
still incomplete.  I view this as distasteful, even just by itself.  It is
not obvious to me that the RDFS entailment is false.  (See below for a vital
caveat on this.) However, I have not yet had a chance to investigate the
situation in as much depth as I would like so I am not prepared to indicate
that I view the RDFS entailment lemma as true.

I believe that my messages concerning this issue have not gone beyond the
incompleteness of the RDF closure (now entailment) rules.  I further
believe that the known problems going into last call and the significant
post-last-call changes require that the RDF Core working group adequately
address new comments on their documents and not limit simply address a
narrow reading of the comments received during that last-call review
period.

I categorically reject the view that changes to one of the RDF documents
must be considered in isolation to determine whether a comment has been
adequately or satisfactorily addressed and that other aspects of the RDF
specification must be excluded from this determination.



The RDFS (and RDF) semantic conditions also depend on the treatment of XML
literals in RDF.  This treatment has recently been changed in a significant
manner that affects both the completeness of the RDFS entailment rules and
the correctness of the RDFS entailment lemma.  The definition of XML
literals in RDF is in RDF Concepts, and is significantly flawed in the
current editor's draft (of 28 July 2003).  Some of these flaws have
reasonably obvious fixes, but others do not, and indeed appear to be the
subject of vigorous debate.  The changes needed to fix the treatment of XML
literals in RDF may affect the completeness of the RDFS entailment rules
and the correctness of the RDFS entailment lemma.

My reading of the treatment of RDF XML literals in RDF Concepts is that
several lexical strings can map into the same value.  Under this reading,
the RDFS entailment lemma is false because it does not recognize different
XML literals that denote the same value.  In any case, the situation with
respect to the completeness of the RDFS entailment rules and the
correctness of the RDFS entailment lemma cannot be determined until the
treatment of RDF XML literals is fixed and finalized.



So, I do not consider that my comment has been adequately addressed.  


Peter F. Patel-Schneider
Bell Labs Research
Lucent Technologies




From: pat hayes <phayes@ihmc.us>
Subject: [proposed close: pfps-05]
Date: Sat, 9 Aug 2003 21:38:45 -0700

> Peter;
> 
> In your comment
> 
> http://lists.w3.org/Archives/Public/www-rdf-comments/2003JanMar/0090.html
> 
> archived as pfps-05 and accepted by the WG:
> 
> http://www.w3.org/2001/sw/RDFCore/20030123-issues/#pfps-05
> 
> you indicated that the closure rules for RDFS in the LC version of 
> the semantics document were incomplete, giving an example inference. 
> The WG accepted this comment and made various changes to the RDFS 
> (and other) rules, and indeed to the RDFS semantics, in order to 
> establish completeness of a relatively simple rule set for RDFS.  You 
> noted that a graph could be RDFS-inconsistent by virtue of the XML 
> datatyping implicit in RDF, and we made further modifications to 
> handle this case.
> 
> Your next response, however
> 
> http://lists.w3.org/Archives/Public/www-rdf-comments/2003JulSep/0178.html
> 
> seems to have enlarged the topic further, to be a kind of general 
> complaint about the style of the document and the way that the rule 
> sets are stated. I have replied to you on this general issue; it may 
> be that we simply disagree about stylistic matters; but in any case, 
> I do not feel that the discussion at this stage is concerned with the 
> topic of the original comment which was accepted by the WG. I note 
> that in your most recent message cited above you use 'incomplete' 
> apparently in a non-technical sense; and that the property of the 
> lemmas which incurs your "deep dissatisfaction"  - that they operate 
> by reducing vocabulary entailment to simple entailment - has been 
> present since the very first draft of the semantics document, and is 
> stated explicitly in the test, but has never been remarked on 
> previously: in particular, none of the comments on the LC version of 
> the semantics document referred to this negatively. I also note that 
> this style of rules has been used directly by implementors apparently 
> with reasonable success with graphs of up to O(10|6) nodes.
> 
> As the formal process is now very late, I must ask you to please 
> indicate whether the changes to the document mentioned in earlier 
> emails are an acceptable response *to the point you raised in your 
> original comment*, viz. the incompleteness, in the sense stated in 
> the document, of the RDFS closure rules (now called entailment 
> rules). This would not, of course, require you to register your 
> satisfaction with every aspect of the document, but it will enable us 
> to proceed with the W3C processes as far as this particular issue is 
> concerned.
> 
> Please Cc your reply to rdf-comments.
> 
> Thanks.
> 
> 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@ihmc.us       http://www.ihmc.us/users/phayes
>

Received on Sunday, 10 August 2003 08:02:41 UTC