WebOnt WG response to RDF Schema Working Draft

		Web Ontology Working Group response to
	 RDF Vocabulary Description Language 1.0: RDF Schema


1.  Executive Summary

This document is a response by the Web Ontology Working Group (WOWG)
to the current RDF Schema working draft [5] published by RDF Core.

In general, our response to this draft is positive; several of the
existing problems in the previous candidate recommendation [2] have
been fixed (including the semantics of repeated rdfs:range and
rdfs:domain constraints and the restriction on cycles in
rdfs:subClassOf and rdfs:subPropertyOf), which brings the semantics of
RDFS into line with those of the Ontology Web Language (OWL) currently
under development by WOWG.

The majority of the issues that we have with this draft are already
covered by the the editorial notes which appear within it. We have
restated certain of the points made in these notes where we feel that
they are particularly important to the development of OWL.

Finally, the explanatory text in this draft is quite sparse and needs
a considerable amount of fleshing out. Without further examples and
cross-references to the relevant sections of other RDF specifications,
it is likely to present a barrier to the understanding of RDF Schema
by novice RDF users.


2.1 Section numbering

While the overall structure of the RDF Schema document is acceptable,
it would benefit from having numbered sections in order to make it
possible to refer to specific parts of the document (this would have
been useful during the preparation of this response!)

2.2 Typographical errors

In Appendix B [5b], the value of the rdfs:comment property on
rdfs:subClassOf is incorrect, and appears to have been cut and pasted
from the comment for rdf:type. It should read "Represents a
specialization relationship between classes of resource", or similar.

This error also appears in the table entitled RDF Properties [5a].

2.3 Changes from previous version

This document is likely to be read before the Model Theory, especially
by novice users. The changes that have been made to the semantics of
multiple rdfs:range and rdfs:domain properties from the previous
version [2] are significant, and should be made explicit in this
document. Similarly, the document should explicitly state that it is
now possible to create cycles in the rdfs:subClassOf and
rdfs:subPropertyOf properties.

2.4 rdf:value

The document notes that rdf:value [5c] is used to identify "the
principal value (usually a string) of a property when the property
value is a structured resource", but there is no explanation of what
is meant by a structured resource. Is this the same as a structured
property value, as mentioned in [3a]? The only existing examples of
the use of rdf:value use are in Model and Syntax [1a,1b].

The RDF Issue Tracking document notes [7a] that the Model Theory [6]
should state that rdf:value has no specific semantics beyond those
ascribed to a property, but it is not mentioned in the MT, nor it is
mentioned in the revised RDF/XML syntax [4].

This text would benefit from further clarification and from
cross-references to the relevant sections of [1],[3] and [6].

2.5 Datatypes

The ability to talk about datatype values is an important requirement
for an ontology language that has been identified by WOWG [8a] and
that appears in several of its open issues [9a,9b]. The DAML+OIL
language, which WOWG has taken as a starting point for its work, takes
an approach to the representation of datatypes which ascribes special
semantics to the rdf:value property [10a] (a potential conflict with
RDF Core's decision regarding rdf:value, as mentioned in 2.4).

The RDF Schema draft makes no mention of datatype issues, nor does it
contain placeholders for references to the RDF Datatypes document
under development by RDF Core. The RDF Schema document should make
clear the relationship between datatypes and schemas; we make this
comment here in anticipation of the publication of the RDF Datatypes
working draft.

2.6 Containers

The section on containers in this document is particularly sparse, and
needs more concrete examples and cross-references to the relevant
sections in the RDF Primer [3] and Model Theory [6], possibly with a
recapitulation of the behaviour of the rdf:li pseudo-property.

In addition, although the rdfs:member property that has been
introduced in this draft appears in the entailments in the Model
Theory [6a], the existing class rdfs:ContainerMembershipProperty does
not. Strictly speaking, this is an issue with the Model Theory, rather
than the RDF Schema draft; we include it here because we believe that
this section needs to be cross-referenced to the Model Theory in order
to clarify the definitions.

2.7 OWL naming

The section on domain and range vocabulary [5d] makes a reference to
"W3C WebOnt language". This should read "W3C Ontology Web Language

2.8 rdfs:isDefinedBy ambiguity

The definition of the rdfs:isDefinedBy property is ambiguous. The
comment in the RDF Schema definition in [5b] states that the object of
this property should be the namespace of the resource which is its
subject, but the definition in the main text [5e] states that it is
used to indicate the resource which defines the subject resource - it
isn't clear whether these definitions are compatible. This has also
been raised as an issue by RDFCore [7b].

3.  References

[1] Resource Description Framework (RDF) Model and Syntax Specification,
    W3C Recommendation, 22 February 1999

    [1a] Qualified Property Values

    [1b] Non-Binary Relations

[2] Resource Description Framework (RDF) Schema Specification 1.0,
    W3C Candidate Recommendation, 27 March 2000

[3] RDF Primer, 
    W3C Working Draft, 26 April 2002

    [3a] Structured Property Values

[4] RDF/XML Syntax Specification (Revised), 
    W3C Working Draft, 25 March 2002

[5] RDF Vocabulary Description Language 1.0: RDF Schema,
    W3C Working Draft, 30 April 2002

    [5a] RDF Properties

    [5b] RDF Schema as RDF/XML

    [5c] rdf:value

    [5d] Domain and Range vocabulary

    [5e] rdfs:isDefinedBy

[6] RDF Model Theory, 
    W3C Working Draft, 29 April 2002

    [6a] RDFS-entailment and RDFS closures

[7] RDF Issue Tracking, 
    2 May 2002

    [7a] Issue rdfms-replace-value: Suggestion that the rdf:value property
         be replaced by rdf:toString 

    [7b] Issue rdfs-isDefinedBy-semantics: Must the value of an
	 rdfs:isDefinedBy property be a schema? 

[8] Requirements for a Web Ontology Language, 
    W3C Working Draft, 7 March 2002

    [8a] Datatypes

[9] Web Ontology Issue Status, 
    W3C Working Draft, 20 May 2002

    [9a] Structured Datatypes
    [9b] Datatypes

[10] DAML+OIL (March 2001) Reference Description,
    W3C Note, 18 December 2001

    [10a] Datatype Values

Nick Gibbins                                            nmg@ecs.soton.ac.uk
Advanced Knowledge Technologies                    tel: +44 (0) 23 80592831
University of Southampton                          fax: +44 (0) 23 80592865

Received on Thursday, 6 June 2002 11:16:10 UTC