Re: question about lexical and value spaces

From: Dave Peterson <davep@iit.edu>
Subject: Re: question about lexical and value spaces
Date: Wed, 9 Jan 2008 17:01:34 -0500

> At 11:41 AM -0500 2008-01-09, Peter F. Patel-Schneider wrote:
> >Hmm.  I hadn't thought of that issue.
> >
> >The kind of datatype that I was interested in was a datatype with
> >uncountably many values, e.g., reals.  In XML Schema 1.0 one could not
> >have such a datatype, but it appears to my reading that they would be
> >allowed in XML Schema 1.1, and I was checking whether this was actually
> >the case.
> 
> That you probably won't be able to do; right now there is no capability
> for anyone other than W3C to add new primitive datatypes.  It's always
> possible that a capability to accept implementation-defined datatypes
> may be provided, but none has been proposed and accepted.  Until/unless
> such a proposal is accepted by the WG, I believe a processor providing
> such a datatype would not be XSDL-compliant.
> 
> I'm curious as to how you would implement such a datatype, and how you
> would identify or share values that have no lexical representation.  It
> sounds nice in theory, but also sounds unimplementable.  I presume storing
> approximations is not adequate; decimal already does that, to any degree
> of approximation that you wish.
> -- 
> Dave Peterson
> 
> davep@iit.edu

Perhaps I misspoke a bit.  The OWL WG is considering adding a primitive
datatype *to the OWL datatypes*.  This would not be an XML Schema
datatype.  However, it would be very nice if the new datatype conformed
to the rules for XML Schema datatypes, which is why I asked my question.

The proposed new datatype would be reals.  The problem with reals as far
as XML Schema 1.0 datatypes go is that not all values would have lexical
forms (obviously).  Implementing such a datatype is, as you say, not
trivial.

If you a reasoning about numbers it is often the case that you don't
need to transmit solutions, just know that a solution exists.  If you do
need to transmit reals, then you do need a lexical representation.  The
final solution may be that the lexical representation is reduced integer
fractions (i.e., the rationals), or it may be that an enhanced lexical
representation is allowed (e.g., adding roots).

peter

Received on Thursday, 10 January 2008 08:18:34 UTC