proposals for the underlying principles of OWL

		Proposals for the W3C Web Ontology Working Group


The following proposals provide resolutions of 
Issue 4.6 EquivalentTo
Issue 5.1 Uniform treatment of literal/data values
Issue 5.3 Semantic Layering
Issue 5.5 List syntax or semantics
Issue 5.9 Malformed DAML+OIL Restrictions
Issue 5.10 DAML+OIL semantics is too weak
Issue 5.12 Entailing inconsistencies
Issue 5.19 Classes as instances

The following issues are also addressed, but can easily be changed.
Issue 5.20 Should OWL provide synonyms for RDF and RDFS objects?
Issue 5.21 drop disjointUnionOf
Issue 5.22 owl:Class still needed

I suggest that these proposals be voted on at the 5 September
teleconference.



Syntax:

1/ The reference syntax for OWL is as in the abstract syntax document,
   at http://www.w3.org/..., with a updated draft version at
   http://www-db.research.bell-labs.com/user/pfps/owl/specification.html
   a) The only OWL knowlege bases are the ones derived from the <ontology>
      production in Section 3.

2/ The exchange syntax for OWL KBs is defined by the transformation from
   the reference syntax in Section 6 of the abstract syntax document.
   a) The only OWL knowledge bases in exchange syntax are ones that are
      transformed versions of OWL knowledge bases.

Semantics:

1/ The normative semantics of OWL knowledge bases is defined in the OWL
   model theory document, with a draft version currently at
   http://www-db.research.bell-labs.com/user/pfps/owl/semantics.html 

2/ An alternative, non-normative, RDFS-compatible semantics is given in
   draft form for OWL KBs that maintain a separation between names for
   classes, properties, and individuals at
   http://www.coginst.uwf.edu/users/phayes/RDFS2OWL-D.html    
   This non-normative semantics is believed to be compatible with the
   normative semantics. 
   a) Only OWL KBs that maintain the name separation can be used in this
      semantics.


How the issues are resolved can only be completely determined by
understanding the various documents.  However, the basics are as follows:

Issue 4.6 EquivalentTo
	EquivalentTo is removed from the language, as it is ill-typed.
Issue 5.1 Uniform treatment of literal/data values	
	There is a strict separation between OWL object and data values.
	Removing the separation has computational consequences.
Issue 5.3 Semantic Layering
	The semantic layering of OWL on top of RDFS is that OWL is a theory
	in an extension of RDFS.  In this theory, the OWL domain of
	discourse is not the entire RDF domain of discourse.
Issue 5.5 List syntax or semantics
	OWL knowledge bases do not abuse list syntax, thus making the
	question somewhat moot.  In the native model theory, lists are
	syntax-only.   In the RDFS layering, lists are outside of
	the OWL domain of discourse. 
Issue 5.9 Malformed DAML+OIL Restrictions
	OWL knowledge bases do not abuse restriction syntax.
Issue 5.10 DAML+OIL semantics is too weak
	The OWL model theory has the desired entailments, as does the RDFS
	layering. 
Issue 5.12 Entailing inconsistencies
	The OWL model theory is a standard model theory in which
	inconsistencies result in the loss of all interpretations.
Issue 5.19 Classes as instances
	OWL knowledge bases can use the same name for classes and
	instances, but this does not identify the class and the instance.
	OWL knowledge bases used in the RDFS layering must use separate
	names for classes and instances.
Issue 5.20 Should OWL provide synonyms for RDF and RDFS objects?
	Synonyms are not provided, but could be.  However, there is no
	mechanism in OWL to state this, so the equivalence would have to be
	via semantic conditions.
Issue 5.21 drop disjointUnionOf
	disjointUnionOf is not in the syntax, but could be added.
Issue 5.22 owl:Class still needed
	owl:Class is different from rdfs:Class and thus is needed.



Peter F. Patel-Schneider
Bell Labs Research

Received on Wednesday, 28 August 2002 12:38:22 UTC