XMLLiteral in OWL

During the last telecon I was actioned with Jeremy and Ian
to look at the problem of XMLLiteral in OWL and propose
a decision [1].
In this message I summarize the problem and the decision
that was already proposed during the meeting.
(See, e.g., [3-7]  for earlier discussion about this problem.)

- The OWL design allows the possibility of doing without 
semantic conditions on XMLLiteral.  This is visible in
the Test document (see tests 201-205).
- S&AS Section 5 (RDF-compatible model-theoretic semantics)
does not allow the possibility of doing without semantic
conditions on XMLLiteral.  OWL Full interpretations as well
as OWL DL interpretations are D-interpretations (from the
RDF Semantics document) and thereby always incorporate
semantic conditions on XMLLiteral.
- This mismatch between S&AS and Test disturbs a bigger
point: as is described in the first paragraph of Test,
S&AS is the primary normative document about OWL, and
Test is a 'subsidiary' document, aiming to give examples
and clarification of S&AS.     As I wrote earlier,
>If the document
>is left with an error like this, how can a reader decide
>for any statement of S&AS whether it is reliable?

Example: the RDF graph
      v p l
      p rdfs:range rdfs:Literal
where l is an ill-typed XML literal, is DL and 
Full-inconsistent according to S&AS Section 5,
whether the datatype map D contains XMLLiteral or not.
Test allows the possibility to not include XMLLiteral
in the datatype map, in which case this RDF graph
becomes consistent for DL or Full.


Proposed solution:

incorporate the three changes precisely described in 
Jeremy's note "possible compromise on rdf:XMLLiteral" [2]
*and* add one sentence to S&AS Section 5 following
the sentence just before Section 5.1:
  "If, however, any conflict should ever arise between
  these two forms, then the Direct Model-Theoretic
  Semantics takes precedence."
The new sentence following this sentence should express
that this applies (so that the direct semantics takes
precedence) when XMLLiteral is not in the datatype map.

(To summarize, the three changes described by Jeremy in [2]
ensure that
-an OWL interpretation (S&AS Section 5.2) always assumes
XMLLiteral in its datatype map (not mentioning this
would misleadingly suggest that this is not necessary)
-the datatype map for OWL Full always includes XMLLiteral
-Test 205 does not apply to OWL Full.)

During the last telecon support for this solution was expressed
by Jeremy and me.


[1] http://lists.w3.org/Archives/Public/www-webont-wg/2004Jan/0040.html
[2] http://lists.w3.org/Archives/Public/www-webont-wg/2003Dec/0100.html

[3] http://lists.w3.org/Archives/Public/www-webont-wg/2003Dec/0035.html
[4] http://lists.w3.org/Archives/Public/www-webont-wg/2003Dec/0042.html
[5] http://lists.w3.org/Archives/Public/www-rdf-comments/2003OctDec/0233.html
[6] http://lists.w3.org/Archives/Public/www-webont-wg/2003Dec/0102.html
[7] http://lists.w3.org/Archives/Public/www-webont-wg/2003Dec/0106.html

Received on Monday, 19 January 2004 10:10:27 UTC