comments on 26 September version of RDF Semantics document

	Comments on RDF Semantics (version dated 26 September 2003)


I did a quick read of the most of RDF Semantics, along with a careful
investigation of a couple pieces.  Here are my comments.

1/ The major new issue I have with the document is the unsupported change
of the entailment rules from informative to normative.  

2/ The removal of the translation to LBase is a major improvement.  I would
have prefered different wording in Section 0.1 concerning this translation
but can live with the new situation.

3/ The document is of several minds with respect to just what the semantics
of RDF (and RDFS) is.  There is wording in Section 1.3 to the effect that
the definition of the semantics of RDF and RDFS excludes the entailment
rules.  How can this coexist with having the entailment rules be normative?

4/ The figures in the copy of the document that I reviewed have naming
problems, but I am told that these have since been fixed.

5/ The Monotonicity and Compactness Lemmas are only stated for simple
entailment.  However, readers could easily be left with the impression that
the other forms of entailment could also have the desirable properties that
accrue from these lemmas.

6/ The characterization of simple entailment rules is misleading as they do
not currently correspond to the instance lemma.  The comment on how to make
this change at the end of Section 7.1 does not adequately serve to correct
the misimpression.

7/ The document needs a better description of consistency and
inconsistency.

8/ The change to the RDFS entailment lemma is incorrectly noted in the
change log as editorial.


9/ I view the development and statement of the entailment lemmas as
fundamentally flawed.  I look for soundness and completeness in any such
syntactic characterization of entailment.  Neither simple entailment, RDF
entailment, nor RDFS entailment have both these characteristics.  


9a/ Section 7.1 does not provide any statement at all on whether the simple
entailment rules are complete with respect to simple entailment.  There is
a statement that the rules are sound, and that an agumentation of them
correspond to the instance lemma, but no proof is presented.

SUGGESTION:  I suggest that Section 7.1 be augmented with the following
Lemma and that its proof be added to Appendix A.

	Simple entailment lemma. S simple-entails E iff there is a graph
	that can be derived from S by the application of the simple
	entailment rules that is a supergraph of E.  (Proof in Appendix A.)

This lemma requires that the simple entailment rules be augmented to allow
vvv in rule se1 and bbb in rule se2 to be a blank node.


9b/ Section 7.2 does not provide an adequate syntactic characterization of
RDF entailment.

SUGGESTION:  I suggest that Section 7.2 be augmented by the following
change to the RDF entailment lemma and that its proof in Appendix A be
modified accordingly.

	RDF entailment lemma.  S rdf-entails E if and only if ... and which
	is a supergraph of E.  (Proof in Appendix A.)


9c/ Section 7.3 does not provide an adequate syntactic characterization of
RDF entailment.

SUGGESTION:  I suggest that Section 7.3 be augmented by adding the
following lemma and providing a proof for it in Appendix A.

	RDFS consistency lemma.  S is rdfs-consistent if and only if it is
	not possible to derive a supergraph of a graph of the following
	form from S plus the RDF and RDFS axiomatic triples by the
	application of the simple, RDF, and RDFS entailment rules

		uuu vvv rrr^^rdfs:XMLLiteral .
		vvv rdfs:range www .
		www rdfs:subClassOf rdfs:Literal .

	where rrr is a non-well formed XML literal.  (Proof in Appendix A.)

I believe that this is a correct characterization of RDFS consistency but
this would have to be shown in the proof.

SUGGESTION:  I suggest that Section 7.3 be augmented by the following
change to the RDFS entailment lemma and that its proof in Appendix A be
modified accordingly.

	RDFS entailment lemma.  If S is rdfs-consistent then S rdfs-entails
	E if and only if ... and which is a supergraph of E.  (Proof in
	Appendix A.)


Peter F. Patel-Schneider

Received on Monday, 29 September 2003 09:49:27 UTC