Re: OWL comment - data ranges (was blank nodes in OWL DL)

Dave,

Thank you for your comments.

I will attempt to provide some further clarification regarding the
WG's decision not to support naming of data ranges. The second part of
your comment (regarding bNodes) is/will be dealt with separately.

The issue of named dataranges was discussed at the editors meeting in
Boston. (see [1]). The following potential problems were identified:

1. Clearly, we would like to have access in OWL to a full range of
user-defined XML Schema datatypes derived from the built-in datatypes
that can already be used in OWL (see [2]). This would include
enumerated datatypes corresponding to OWL dataranges. We expect
XML:Schema to ultimately provide a mechanism to support this. Naming
dataranges in OWL would provide a "competing" mechanism (i.e.,
provide an alternative way to name user defined datatypes), and this
could interact in an undesirable way with the XML:Schema mechanism as
and when it is introduced.

2. OWL DL is designed so as to allow reasoning about datatypes and
values to be cleanly separated from reasoning about classes and
individuals. Introducing OWL names for dataranges may compromise this
design.

It was therefore decided not to include them in the language at
present. It may be possible to add them in the future as and when a
thorough investigation of the issues proves that they would not have
any adverse effects.


Please reply to this message as to whether this response is satisfactory,
copying public-webont-wg@w3.org. Again, thank you for your comments.

Ian Horrocks




On July 23, Dave Reynolds writes:
> 
> Jim,
> 
> Thank you for your response to the Jena team comments on these issues.
> Overall this response is not (yet) acceptable. 
> 
> (a) Issue: Named data ranges
>      Your response: postpone
> 
> We understand that the working group cannot name user-defined XSD datatypes 
> and that matter should be raised with the XML Schema working group.
> 
> Our concern was more one of uniformity - it seems possible to have both 
> named and unnamed classes, why not data ranges? The more uniform a language 
> is, the easier the API and the fewer the support calls.
> 
> As an example could this:
> 
> <owl:DataRange rdf:about="#MyDR">
>     <owl:oneOf>
>        <rdf:List>
>          <rdf:first>foo</rdf:first>
>          <rdf:rest rdf:resource="&rdf;nil"/>
>        </rdf:List>
>     </owl:oneOf>
> </owl:DataRange>
> 
> be included in OWL DL, for greater uniformity with other unnamed things in 
> OWL DL (which can optionally be named).
> 
> I confess to not understanding the research problems that you refer to as 
> being raised by naming data ranges. If there is some non-trivial problem 
> here then we certainly accept this is not a sufficiently important issue to 
> warrant additional research at this stage in the process.
> 
> Has the WG discussed this question?
> None of the three links you gave seemed directly related to our request:
> [1]
> http://www.w3.org/2001/sw/WebOnt/webont-issues.html#I5.8-Datatypes
> [2]
> http://www.w3.org/2001/sw/WebOnt/webont-issues.html#I4.3-Structured-Datatypes
> [3]
> http://lists.w3.org/Archives/Public/www-webont-wg/2002May/0040.html
> 
> 
> (b) Issue: bNodes as object in multiple triples
>      Your response: no change due to lack of working group consensus
> 
> The links you gave were all to the discussion before the last call.
> Please give an indication of the WG discussion of this issue in response
> to last call comments.
> 
> Specifically you said:
> [[
> The WG was concerned that the handling of blank nodes has not yet been 
> shown to be able to be handled in the correspondence proof of Appendix A of 
> the Semantic Document [6].
> ]]
> However the following messages claim to provide such a proof:
> http://lists.w3.org/Archives/Public/www-webont-wg/2003Jun/0294
> http://lists.w3.org/Archives/Public/www-webont-wg/2003Jun/0017
> 
> If this proof is indeed in error an analysis demonstrating the error could 
> supply the test case that we requested.
>    [[
>    A rationale for not permitting this in OWL DL
>    should be given, preferably as a test case in OWL Full
>    showing an OWL Full non-entailment that would hold in
>    OWL DL if such triples were permitted.
>    ]]
> 
> 
> (c) Issue: cycles of bNodes
>      Your response: postpone
> 
> Your follow up response was disappointing but acceptable.
> http://lists.w3.org/Archives/Public/public-webont-comments/2003Jul/0045
> 
> Dave, for the Jena team
> 
> 
> Jim Hendler wrote:
> > 
> > Dave-
> >   thanks for all the comments you raised on the
> > public-webont-comments@w3.org list.  This is the response to the
> > remaining one:
> > 
> > >This is a comment about blank nodes in OWL DL on behalf of the Jena team.
> > 
> > >   - We would like it to be possible to name dataranges
> > 
> > The Web Ontology WG discussed this issue and others to do with naming
> > various datatypes under the issue "datatypes" which is issue 5.8 on
> > our issues list [1] and in our issue 4.3 of "structured datatypes"
> > [2], which included a discussion of naming datatranges raised by Ziv
> > Hellman [3]
> > Summarizing what can be seen in the discussions referenced there, the
> > WG felt that
> >    (i) we should yield to RDF Core's decisions on datatyping, and
> >    (ii)that we shouldn't create new URIs for datatypes that might
> > conflict with an eventual decision in the handling of XML Schema
> > Datatypes by the XML Schema group.
> >   In light of these considerations, the group decided to postpone
> > issue 4.3 and not add named dataranges at this time.
> >   The above assumes your question is specifically about complex
> > datatypes and ranges (i.e. being able to say a teenager has as age
> > property value between 13 and 19).  Unamed datatranges could also be
> > asked about the ability to create datatypeProperties that are created
> > in a manner corresponding to the creation of unnamed
> > ObjecttypeProperties.  This was not identified by the WG as a
> > requirement and raises some issues of a research nature as to the
> > handling of these within the DL restrictions.
> > 
> > - It is natural in Jena to permit the java objects
> >    representing blank nodes which
> >    represent descriptions and restrictions to be
> >    used more than once, and hence as the object of
> >    more than one triple.
> > 
> >    A rationale for not permitting this in OWL DL
> >    should be given, preferably as a test case in OWL Full
> >    showing an OWL Full non-entailment that would hold in
> >    OWL DL if such triples were permitted.
> >    Below is an example such a test case, which we believe is an
> >    OWL Full entailment. If on the contrary it were a
> >    non-entailment, that would satisfy this comment.
> > 
> > The issue of the proper treatment of blank nodes is a complex one,
> > and has received a great deal of discussion on the WG.  Issue 5.26
> > [4] involves many subissues involving the mapping between RDF and
> > OWL.  Part of that discussion involved the treatment of blank nodes
> > similar to your discussion above.  The group was unable to reach
> > consensus on this issue, and in fact ended up closing the issue over
> > the objection of Jeremy Carroll of Hewlett-Packard, his objection can
> > be seen in [5].  The WG was concerned that the handling of blank
> > nodes has not yet been shown to be able to be handled in the
> > correspondence proof of Appendix A of the Semantic Document [6].
> > 
> > - The RDF support within Jena permits users to:
> >    - use annonymous nodes as the object of
> >      more than one triple
> >    - have cycles of anonymous nodes
> >    While the syntactic restrictions between, for example,
> >    unnamed individuals and descriptions are understandable,
> >    it is not clear why OWL DL has not permitted, for example,
> >    an unnamed individual to be the object of more than
> >    one triple, or an unnamed individual to be the object
> >    of a triple of which it is the subject.
> >    Please either relax this constraint or offer a rationale.
> > 
> > I am pasting in here the response that Peter Patel-Schneider sent to
> > another comment raising this same issue - his answer can be seen in
> > full in [7]
> > 
> > The constraints on owl:equivalentClass triples can be articulated in terms
> > of Hamilton paths in the component graphs that are created by considering
> > only connected groups of blank nodes and named nodes that are connected to
> > these blank nodes.  Neverthless, this is expensive to implement.
> > 
> > The working group decided on 29 May 2003, as recorded in
> > http://lists.w3.org/Archives/Public/www-webont-wg/2003May/0402.html, to
> > change the mapping for owl:equivalentClass from paths to connected graphs.
> > This should be much easier to implement.
> > 
> > This change is reflected in the editor's draft of S&AS as of 30 May 2003,
> > which says:
> > 
> > --------
> > S: EquivalentClasses(description1 ˆÉ¬ÇˆÇ¬Ö descriptionn)
> > 
> > T(S): T(descriptioni) owl:equivalentTo T(descriptionj) .
> >         for all <i,j> in G where G is a set of pairs over
> >         {1,...,n}x{1,...,n} that if interpreted as an undirected graph
> >         forms a connected graph for --------
> > 
> > If you'd like to review this in context, you can take a look at the
> > editor's draft, in the the "Transformation to Triples" table at
> > http://www-db.research.bell-labs.com/user/pfps/owl/semantics/mapping.html
> > 
> > I hope these responses will demonstrate the rationale behind the
> > decisions the WG has made on these issues.  We hope you can accept
> > our postponement of one, reluctance to change our response on
> > another, and changes made to fix the third.
> >   -Jim Hendler
> >   for WOWG
> > 
> > [1] http://www.w3.org/2001/sw/WebOnt/webont-issues.html#I5.8-Datatypes
> > [2]
> > http://www.w3.org/2001/sw/WebOnt/webont-issues.html#I4.3-Structured-Datatypes
> > [3] http://lists.w3.org/Archives/Public/www-webont-wg/2002May/0040.html
> > [4]
> > http://www.w3.org/2001/sw/WebOnt/webont-issues.html#I5.26-OWLDLSyntax
> > [5] http://lists.w3.org/Archives/Public/www-webont-wg/2003Mar/0264.html
> > [6]
> > http://www.w3.org/TR/owl-semantics/proofs.html#A.1
> > [7]
> > http://lists.w3.org/Archives/Public/public-webont-comments/2003Jun/0022.html
> > 
> > --
> > Professor James Hendler                           hendler@cs.umd.edu
> > Director, Semantic Web and Agent Technologies     301-405-2696
> > Maryland Information and Network Dynamics Lab.    301-405-6707 (Fax)
> > Univ of Maryland, College Park, MD 20742          *** 240-277-3388 (Cell)
> > http://www.cs.umd.edu/users/hendler      *** NOTE CHANGED CELL NUMBER ***
>

Received on Sunday, 27 July 2003 17:13:50 UTC