W3C home > Mailing lists > Public > www-webont-wg@w3.org > July 2002

Re: OWL Reference Description for F2F 3

From: Peter F. Patel-Schneider <pfps@research.bell-labs.com>
Date: Mon, 15 Jul 2002 08:02:52 -0400
To: mdean@bbn.com
Cc: www-webont-wg@w3.org
Message-Id: <20020715080252R.pfps@research.bell-labs.com>

		Comments on 
		OWL Web Ontology Language 1.0 Reference Description
		Version of 2002/06/17

Problems:

- Some parts of the document are missing, in particular Appendix C.  Adding
  them is not simply a matter of editorial work.  Without them I cannot
  complete a review of the document.

- The term ``instance'' is not defined in the document, even though it is
  heavily used in the document (with two different meanings).

- The document refers to many other documents whose status is unclear, and
  will probably not exist for some time yet.  The document should make the
  status of these documents clear.  Such documents include
  - the WebOnt WG guide document
  - the WebOnt WG formal specification document
  - the various WebOnt WG documents on non-normative syntaxes
  - a model-theoretic semantics for OWL
  - an axiomatization for OWL

- The pointers to issues is incomplete.  In particular, resolution of the
  layering issue and the too-weak-semantics issue may cause changes to the
  fundamental structure of OWL, resulting in changes to portions of the
  introductory remarks as well as other changes throughout the document.

- Some of the examples do not appear to exist yet, such as OWL-ex.owl.
  Without these examples portions of the document cannot be verified.

- The document should talk about RDF/XML (or is it XML/RDF?).

- The form of an OWL document is not specified in the document.  Is an OWL
  ontology an RDF/XML description?  Or is an OWL ontology more akin to an
  RDF/XML document?  In particular, can an OWL document have multiple OWL
  ontologies in it?  What happens if an OWL document contains multiple
  headers?  Are ontology references URIs or URI references?

- Having two different enumeration elements for a class is more complex
  than mentioned in the document.  Perhaps the warning should be modified
  to something like ``Having more than one enumeration for a class has
  unusual effects, and should be avoided.''  Similarly the effect of more
  than one boolean combination is complex, as the effect of both a boolean
  combination and an enumeration.  (Well actually the effect is quite
  simple to state - the boolean combinations and enumerations become
  equivalent - but the ramifications of this are rather complex.)

- Boolean combinations of class elements probably have to be enclosed in
  owl:Class tags.  (This would be a change from the DAML+OIL reference
  document, but I believe that it is wrong there.)

- owl:collection is invalid RDF syntax, as has been pointed out by the RDF
  Core WG.  The RDF Core WG may be working towards providing such a
  facility, but I do not believe that they have reached a resolution on
  this issue that provides for owl:collection.

- The pointer to issue 5.9 is not in the appropriate place.  Issue 5.9
  refers (mostly) to restrictions that have too many or too few
  components, not restrictions that are neither datatype restrictions nor
  object restrictions.

- The effect of multiple restrictions listed as part of a single
  restriction is *not* as stated in the document.  (This would be a better
  place to put the pointer to issue 5.9.) 

- The term ``list'' is used but not defined.

- The lexical representation of XML Schema datatypes is not consistent with
  any proposal for RDF datatyping.  This should be mentioned in the
  document.  The example given is not valid XML.


Minor problems:

- The pointers to issues 5.1 and 5.9 is not correct.

- XML Schema datatypes are referenced from within RDF by URI references,
  not URIs.  Similarly for other things, like class names.

- The warnings about RDFS limitations could be removed, as the RDF Core WG
  has resolved to remove the limitations.

- The notice about restrictedBy could be removed.

- It would probably be better to call non-datatype OWL properties object
  properties. This would affect several places in the document.  (Some
  places already use object property.)

- The ``at least one'' in owl:hasValue is rather odd.  How can an object be
  related to another object more than once? 


Editorial problems:

- The list under class expressions has two ``or''s.

- There is no pointer to the cardinality syntax note.

- Some URI references are missing owl:.

- owl:sameIndividual should be owl:sameIndividualAs

- UnambigousProperty should be UnambiguousProperty


Points raised by the document:

- How will the syntax of OWL be defined?  For example, the document says
  that an owl:complementOf element contains a single class expression, but
  this is not expressible in RDF Schema.  Will any collection of RDF
  triples constitute an OWL KB?
Received on Monday, 15 July 2002 08:03:23 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:51 GMT