W3C home > Mailing lists > Public > w3c-rdfcore-wg@w3.org > November 2002

Re: xsd:float and xsd:decimal

From: pat hayes <phayes@ai.uwf.edu>
Date: Sat, 30 Nov 2002 21:57:20 -0600
Message-Id: <p05111b0bba0f0cede169@[]>
To: Dan Connolly <connolly@w3.org>
Cc: w3c-rdfcore-wg@w3.org

>On Fri, 2002-11-22 at 13:06, Brian McBride wrote:
>>  At 11:52 22/11/2002 -0600, Dan Connolly wrote:
>>  >Please include this in the test stuite.
>>  Lets not.
>>  Why are we creating such dependencies between RDF and xml schema.
>Umm... because we're chartered to?
>"RDF Schema must use and build upon XML Schema datatypes
>to the fullest extent that is practical and appropriate."
>   -- http://www.w3.org/2001/sw/RDFCoreWGCharter
>>   The RDF
>>  test cases should test rdf.  Lets keep to any dependencies of subtle
>>  aspects of xml schema datatypes out of it.
>On the contrary! Let's be absolutely clear about which
>parts of the schema spec we depend on and which we don't.

We can do that without committing ourselves to particular details. 
What RDF needs to know about a set of datatypes is entirely concerned 
with identities between values, as Jeremy noted. That is, it needs to 
know all equalities of the form

<lexical-form-1>^^dtype1 = <lexical-form-2>^^dtype2  (where dtype1 
may be the same as dtype2, or not)

and it needs to know any subclass relationships, ie when the set of 
all values of one datatype is a subset of the set of all values of 
another. This latter needs to said separately since it would be an 
infinite number of equations, in general.

If any, or any part, of this information is missing, then an RDF 
engine will not be able to draw as many conclusions as it could if 
they weren't. In particular, it would be fine to just have some 
equations where the datatypes were the same, which would be a 
commonly available subset of all information: it would be the 
'intra-datatype' information without the 'inter-datatype' information.

There might be more information available, as well, eg the values of 
various properties, as in JOs' examples. Making use of information 
like this goes outside RDF but could be defined to be a semantic 
extension. What more can we say, in general?


>This isn't a question about how XML Schema works;
>it's about how our datatypes design integrates
>with XML Schema.
>If Jeremy and I have not convinced the WG that this is how
>XML Schema works, questions from those who doubt it
>should "be referred to the authorities
>who defined the datatypes", i.e. www-xml-schema-comments.

You have not convinced me that this is how XML Schema is intended to 
work. The other authorities have not convinced me that it is not, I 
hasten to add. I think that there is confusion within the XML spec on 
the matter, and that the answers are unclear, which is why it is 
particularly important that we do not make pronouncements that might 
be made false by the other spec writers. And in any case I agree with 
Brian that these questions are none of our business. What we should 
say is that IF the spec of some datatypes says that the items in one 
value space are a subset of those in another, THEN the relevant 
rdfs:subClassOf property holds between the datatype classes in any 
datatype interpretation. We can say that, sure: but we should not, in 
my view, actually assert that any such relationships hold between 
particular XSD datatypes. Those questions are not for us to decide, 
and even if they are true now, they might not be in the future.

>If the WG doesn't agree that this is how our datatypes
>design works with XML Schema, somebody will please
>explain to me how it does work, preferably with
>an explanation that includes a few test cases.
>>   This point seems to be subtle
>>  enough to have misled DanC.
>That argues that it deserves a test case, no?
>>   And lets not give ourselves any more work than
>>  we have to.  We are a little short of time after all.
>If we don't have time to do this design work, let's please
>re-open the datatypes decision and re-consider rdfs:format
>This rdfs:datatype design is complex; it deserves lots
>of test cases. I'm not interested in doing it
>half way.
>>  Dan, if you want this to go into an rdf test suite, I'd like to see a
>  > compelling reason why its necessary for RDF.
>First, to be clear, by RDF, I presume you mean to include
>RDFS and datatypes; I don't claim this test is
>necessary for RDF per se.
>Then, as I say: I think this rdfs:datatype design is complex;
>I can't support a request to take this design to Proposed
>Recommendation until I have seen enough implementation
>experience (i.e. test cases) to validate the design.
>Perhaps that's not compelling.
>And if push came to shove, I'd accept that a
>the base64binary/hexBinary test case I gave
>actually tests the same thing that this
>test case does.
>But when it comes to test cases, the more
>the merrier, no?
>>   The XML schema guys can do
>>  their own.
>I am an XML Schema guy; please let's not make this
>an us-versus-them thing.

Its not a case of 'versus', but of each spec being careful to 
restrict itself to defining what it sets out to define and to defer 
to other specs matters which are within their purview. That just 
seems like good cooperation policy to me: it gives each group rights 
to make proclamations in its territory. I wouldn't want XML Schema to 
be making claims about what rdfs:Datatype was supposed to mean, and I 
don't see why we need to say anything about what xsd:integer is 
supposed to mean.


IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola              			(850)202 4440   fax
FL 32501           				(850)291 0667    cell
phayes@ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
s.pam@ai.uwf.edu   for spam
Received on Saturday, 30 November 2002 23:53:30 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 20:24:18 UTC