issues to be resolved before last call

		Technical Issues to be resolved before last call


Here are some issues that I feel must be resolved before OWL goes to last
call.  I have battled to get the RDF issues resoved, but to little
success.  I would like to have formal votes on all the proposed resolutions
herein.



Issues with XML Schema:

1/ The value spaces of xsd:decimal and xsd:float are not disjoint.  Right
   now in both XSD datatyped RDF and OWL the denotation "234.5"^^xsd:float
   is an element of the class extension of the denotation of xsd:decimal.

   PROPOSED RESOLUTION:  The class extensions of xsd:decimal and xsd:float
   are not disjoint.   The class extension of xsd:float is a subset of the
   class extension of xsd:double.


Issues with RDF Concepts:

1/ The notion of social meaning has no place in the specification of a
   formal system.

   PROPOSED RESOLUTION:  Social meaning, as defined in the RDF Concepts
   document, has no effect whatsoever on the meaning of OWL ontologies.

   NB:  I view this as an extremely serious issue.


Issues with the RDF model theory:

Caveat: The RDF Semantics document may change, and has recently changed.
The issues here may be affected by any upcoming change, and have been
affected by the recent changes, which I have not completely analyzed.  I
have not yet received responses to many of my recent comments on the RDF
model theory nor have I been given any sort of change list for any of the
recent changes to RDF Semantics.  The following is only my best guess at
the current situation.

1/ The class extension of rdfs:Literal does not necessarily include
   strings.  Right now, "a" rdf:type rdfs:Literal . is not RDFS-entailed by
   the empty graph.  Also, without the following fix, "a" rdfs:type _:x
   . _:x owl:complementOf rdfs:Literal . is not OWL Full entailed by the
   empty graph. 

   PROPOSED RESOLUTION:  Add to the second table in the OWL model theory

	if E is 	then CEXTI(SI(E)) =
	rdfs:Literal	LV

   As well, add a note to the effect that this is a change to the meaning
   of rdfs:Literal.

   NB:  This condition *is* already present.  However, I feel that the
        WebOnt working group should explicitly resolve this point.

   NB:  This is a significant extension of the RDFS model theory, with lots
        of observable consequences.

2/ XSD-interpretations in RDF include the problematic XML Schema
   datatypes.  

   Current situation: OWL interpretations are defined from
   D-interpretations where D consists of the appropriate XML Schema
   datatypes, not from XSD-interpretations.

   PROPOSED RESOLUTION:  OWL interpretations are not defined from
   XSD-interpretations. 

3/ There is no definition of literal strings or language tags in the RDF
   Schema document.

   PROPOSED RESOLUTION:  Literal strings are Unicode strings.

   PROPOSED RESOLUTION:  Language tags are as in RDF Concepts.

4/ There is contradictory and unclear information with respect to what is a
   datatype as well as to the treatment of typed literals.  Datatypes are
   sometimes elements of the semantic domain, denoted by URI references,
   and sometimes the URI references themselves.  There is no necessary
   connection between the intended URI reference for a datatype and the
   datatype itself.

   PROPOSED RESOLUTION:  A datatype is a formal object, consisting of 
   a URI reference, a lexical space (which is a subset of the Unicode
   strings), a value space, and a lexical-to-value mapping from the lexical
   space into (not onto) the value space.  In a
   D-interpretation containing the datatype d = <U,L,V,L2V>, 
   a/ D <= ICEXT(I(rdfs:Datatype))
   a/ IS(U) = d;
   b/ ICEXT(d) = V;
   c/ for l in L and t a langauge tag, IL("l"^^U) = L2V(l) and 
		IL("l"@t^^U) = L2V(l)
   d/ for l not in L and t a language tag, IL("l"^^U) is not in LV and
 		IL("l"@t^^U) is not in LV 

   NB:  Under this resolution rdf:XMLLiteral is *not* a datatype.

   NB:  Under this resolution I("aa"@en^^xsd:decimal) is not necessarily
	the same as I("aa"@fr^^xsd:decimal).

5/ The non-normative parts of the RDF model theory, including the closure
   rules and the translation to Lbase, have many errors.

   PROPOSED RESOLUTION: The W3C Web Ontology Working Group notes that the
   RDF Semantics document has numerous errors in its non-normative
   sections.  None of these sections are to be used to provide guidance as
   to the meaning of RDF or OWL where the normative sections of the RDF
   Semantics document are unclear, contradictory, or silent.

   NB: I feel very strongly that WebOnt needs this resolution.  There are
       important places, such as the definition of what is a datatype and
       the treatment of rdf:XMLLiteral, that are unclear or contradictory
       in the RDF Semantics document, but where the non-normative sections
       of the document clearly use only one of the possible
       interpretations. 


Peter F. Patel-Schneider
Bell Labs Research
Lucent Technologies

Received on Friday, 17 January 2003 07:44:56 UTC