- From: <herman.ter.horst@philips.com>
- Date: Mon, 15 Dec 2003 14:59:24 +0100
- To: Jeremy Carroll <jjc@hpl.hp.com>, www-webont-wg@w3.org
- Cc: connolly@w3.org, pfps@research.bell-labs.com, pat hayes <phayes@ihmc.us>, sandro@w3.org
> > >If at some point we were to change the design in the way that Herman apperas >to want (or perhaps to change the understanding of the design reflected in >the test document), then the tests misc-201 thru to 205 would need to be >examined. We would then obsolete #205 whose manifest indicates that the test >is not valid when XMLLiteral is a supported datatype.#204 is identical except >giving the opposite result when XMLLiteral is supported i.e. the test >document makes it clear that this is a substantive issue. > > == > >I agree with Pat's comments about timeliness ... I agree, in addition, with Pat's comments that it seems to be necessary to correct the test cases to become consistent with S&AS [1]. > >While Herman is undoubtedly correct to indicate that this is a logical wart on >the Semantic Web docs it is not a disaster. > >The indended meaning of the RDF docs is clear > >The intended meaning of the OWL docs is clear The phrase 'intended meaning' is not completely operational. If we replace 'intended' by 'expressed', then I agree, and would add that there is inconsistency between S&AS and OWL Test, with respect to XMLLiteral. > >I note that the following implementation sketch describes an implementation >that is both RDF conformant and OWL conformant, and exercises the parts of >the specs that Herman is worrying about. What would be the place of such an implementation sketch? Isn't it the task of Webont to produce consistent specs? > >I gloss over whether this is LIte, DL or Full - minor changes would need to >tbe made ... > >1) the documentation states that the supported datatypes are xsd:integer and >xsd:string. > >2) On input any typed literal of type rdf:XMLLiteral is verified. It is >checked that the lexical form is in the lexical space (this is a no op for >rdf:parseType="Literal", a simple way to verify this for ohter input is to >synthesis an RDF/XML document with hopefully a single rdf:parseType="Literal" >triple in, and parse it - Jena includes working code using this algorithm) > >3) Any rdf:XMLLiteral literal that is not in the lexical space is rejected >(not quite sure what happens, detail) > >4) a complete OWL reasoner supporting the given datatypes is used. > >5) The reasoner can find all RDF entailments which follow from >rdf:XMLLiteral's which have the same lexical form, but cannot conclude that >two with different lexical forms are different. Since no RDFS entailments >follow from this (such reasoning is part of OWL), the reasoner does find all >RDFS entailments > >6) by hypothesis the reasoner finds all OWL entailments > > > >I believe such an implementation could justifiably claim RDFS conformance and >OWL conformance. > >=== > >While there is a blemish in the mismatch between the semantics docs it is not >a showstopper. In my opinion Herman is making too much of this. I would like to point out that there is no inconsistency between RDF Semantics and OWL S&AS. There exists an inconsistency between S&AS and OWL Test. As you noted, tests 201 to 205 would need reconsideration. > >(the observation about datatype theory/map is a plausible editorial change to >test) > > >Jeremy > > Herman [1] http://lists.w3.org/Archives/Public/www-rdf-comments/2003OctDec/0236.html
Received on Monday, 15 December 2003 09:00:07 UTC