Re: Datatyping literals: question and test cases

[Patrick Stickler, Nokia/Finland, (+358 40) 801 9690, patrick.stickler@nokia.com]


----- Original Message ----- 
From: "ext pat hayes" <phayes@ai.uwf.edu>
To: "Jeremy Carroll" <jjc@hpl.hp.com>
Cc: <w3c-rdfcore-wg@w3.org>
Sent: 30 October, 2002 20:12
Subject: RE: Datatyping literals: question and test cases


> I would suggest that we let the model theory merge the cases back 
> together, by ignoring the lang tag for non-XML literals. It has no 
> meaning outside XML in any case: the *strings* "chat"eng and "chat"fr 
> are the same string. If we want to distinguish them, then we should 
> call them 'words' and have a domain of words in the model theory for 
> them to denote.

I like the consistency of this. That all literals are typed literals
and the lang tag is not relevant to the L2V mapping, but is still
present as "syntactic residue" from the RDF/XML for applications
which wish to filter nodes based on that.

And in the future, when RDF provides mechanisms for scoping of
assertions, we can remove and deprecate the lang tags without
impact to the semantics, using the scoping mechanisms in their
place.

I would myself go so far as to also ignore the lang tag for XML
literals, since (a) since it's XML, it can specify its own xml:lang
attribute as required and (b) since XML literals are fragments which
may have application/interpretation in isolation from the RDF/XML
serialization, it has never seemed reasonable to me for xml:lang
specified for the RDF/XML instance to "infect" these fragments.

Patrick

Received on Thursday, 31 October 2002 02:12:52 UTC