W3C home > Mailing lists > Public > www-rdf-comments@w3.org > April to June 2003

Re: pfps-08 last call comment on typed literals

From: Peter F. Patel-Schneider <pfps@research.bell-labs.com>
Date: Fri, 02 May 2003 14:54:49 -0400 (EDT)
Message-Id: <20030502.145449.101252182.pfps@research.bell-labs.com>
To: phayes@ai.uwf.edu
Cc: www-rdf-comments@w3.org

From: pat hayes <phayes@ai.uwf.edu>
Subject: Re: pfps-08 last call comment on typed literals
Date: Thu, 1 May 2003 11:20:48 -0500

> >From: pat hayes <phayes@ai.uwf.edu>
> >Subject: Re: pfps-08 last call comment on typed literals
> >Date: Tue, 29 Apr 2003 16:51:43 -0500
> >
> >>  Peter,
> >>
> >>  In
> >>
> >>  http://lists.w3.org/Archives/Public/www-rdf-comments/2003JanMar/0087.html
> >>
> >>  you raised a last call comment on the RDFCore WD's which was recorded as:
> >>
> >>  http://www.w3.org/2001/sw/RDFCore/20030123-issues/#pfps-8
> >>
> >>  The WG has previously decided to reject this comment:
> >>
> >>  http://lists.w3.org/Archives/Public/www-rdf-comments/2003JanMar/0537.html
> >>
> >>  However, subsequent work has suggested that we reconsider this decision.
> >>
> >>  The treatment of XML literals currently proposed would retain the
> >>  'simple' version (in which the literals are treated as a special
> >>  lexical form) in RDF and RDFS interpretations, but treat
> >>  rdf:XMLliteral as denoting a datatype object in D-interpretations.
> >>
> >>  This would support the entailment you refer to in all datatyped
> >>  interpretations, for typed literals which do not contain language
> >>  tags.  It would not, however, support an inference of the following
> >>  form (in Ntriples):
> >>
> >>  ex:bar owl:sameIndividualAs rdf:XMLLiteral .
> >>  ex:s ex:p "foo"@tag^^rdf:XMLLiteral .
> >>  |-
> >>  ex:s ex:p "foo"@tag^^ex:bar
> >>
> >>  since the RDF semantic conditions require that language tags are
> >>  ignored in non-XML typed literals.
> >>
> >>  Please let us know whether this would be acceptable.
> >>
> >>  Pat Hayes
> >
> >I do not view this as a satisfactory solution to my issue.
> >
> >I do not even view it as in any way better than the previous state of
> >affairs.
> OK; but that is rather a strong rejection, which leaves me rather at 
> a loss how to proceed. In earlier correspondence (off-list) you 
> seemed to indicate that you found the current treatment of datatypes 
> satisfactory; 

The treatment of datatypes now appears to be (relatively) satisfactory with
the sole exception of rdf:XMLLiteral.  (I would still like to have some way
of getting "15" to denote 15 based on range information or something
similar, but I realize that this is not really possible without some
violation of the RDF vision.)

> admittedly at that time our attention had not been 
> drawn to the above case.  There seem to be many ways that OWL could 
> accommodate to this situation. For example, I gather than Webont is 
> likely to recommend that rdf:XMLLiteral be deprecated in OWL. If so, 
> it would be harmless for OWL to deprecate, or even forbid, the use of 
> lang tags in literals (since they play no role anywhere but for 
> XMLLiteral) , and then the above would seem to solve the problem you 
> raised in your original comment.  Another, less draconian, solution 
> would be to forbid any OWL assertion which equates anything to 
> rdf:XMLliteral (which, because of its special role in the RDF syntax, 
> has to be treated as an opaque identifier rather than a simple name.) 
> Such equations would be prohibited in OWL-DL and OWL-Lite, in any 
> case.  Your strong rejection, therefore, leads me to believe that 
> there might be other problems than the ones I had diagnosed from your 
> original comment.
> Could you indicate which aspects of this are you are unsatisfied 
> with?  In particular, if the treatment of lang tags were made uniform 
> across all typed literals, so as to support substitution of equals 
> anywhere in the RDF literal syntax in datatyped interpretations, 
> would that be sufficient? There are several ways we could try to 
> arrange that to happen, but it would save some work if you could tell 
> us what your criteria for success would be, so that we can avoid 
> unsatisfactory designs by a process more efficient than Darwinian 
> selection.  Or would you insist on XML literal typing being done in a 
> semantically uniform way between undatatyped and datatyped RDF 
> interpretations? I do not think we can accommodate the latter 
> requirement, so if that is what you find unsatisfactory then there 
> may be no point in attempting to satisfy you on this particular issue.
> Pat

The problem that I see is that there is still this datatype
(rdf:XMLLiteral) that doesn't act like a datatype.

If there was some different syntax for this purpose, then the situation
would be much better.  For example, if literals in triples looked like


then it would be obvious that string literals and XML literals are
different from typed literals.  This would not require any change in
RDF/XML by the way.

I would find it still better if language tags were eliminated entirely.
This would allow completely uniform treatment of literals, with, perhaps,
untyped literals going back to being underspecified.  

Note that even this would not eliminate all the concerns I have with

Peter F. Patel-Schneider
Bell Labs Research
Lucent Technologies
Received on Friday, 2 May 2003 14:55:05 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:15:20 UTC