Re: RDF should allow XML datatypes

_____________Original message ____________
Subject:	Re: RDF should allow XML datatypes 
Sender:	ext Dave Beckett <dave.beckett@bristol.ac.uk>
Date:		Wed, 30 Oct 2002 09:07:31 +0300


> 1. There is no compelling reason to prohibit this given the current RDF
> datatyping solution for which this is a minor modification to the syntax.

As editor of the RDF/XML spec, I feel this is not a minor syntax
change and in particular does not match our abstract syntax for
datatyped literals, so would require a change to RDF.

	In all fairness, Dave, this really isn't a big change to the XML syntax
	and the abstract syntax predicts it symmetrically, but is now
	asymetrical by excluding typed xml literals. I don't consider this
	to be a reasonable argument for not providing the requested
	functionality.
 
> 2. Allowing this will be very useful for OWL which needs to deal with
> structured datatypes

... Plus you can still do it with the rdf:datatype, since
it allows any lexical form to be given as a string, that includes XML
infosets serialised to a string.

	Well, whether a typed literal is serialized as a string or as XML
	is simply a syntactic and convenience issue. If you have the
	URI of a complex datatype such that you can express a typed
	literal, and the lexical form is valid XML, I see no reason whatsoever
	why you shouldn't be able to also express it as a typed XML
	literal. The only effect the XML flag has is to indicate whether
	the lexical form is XML escaped or 'raw' and has no semantic
	significance. In fact, the following entailment should hold:

	some:thing some:prop "&lt;x&gt;foo&lt;/x&gt;"^^some:x .

	entails

	some:thing some:prop xml"<x>foo</x>"^^some:x .

	There is no semantic difference. Just a choice of whether
	or not 'raw' XML is used as a convenience in the RDF/XML
	or not. That XML flag is not so much saying that the lexical
	form is valid  XML (the datatype tells us that) but that it simply
	is non-escaped XML. 

	And, by the way, RDF is datatyping framework agnostic, and
	thus even if XML Schema does not (yet) provide URIs for all
	complex types, that doesn't mean that URIs will not exist for
	complex XML datatypes.

	I agree with and support the proposal for typed XML literals. It's
	straightforward (IMO trivial) to provide, intuitively correct, useful,
	and makes the datatyping solution complete and consistent
	for all lexical forms, whether they are valid XML or not. It's the
	last step towards ceasing to treat XML literals specially or 
	differently in the abstract model.

	Cheers,

	Patrick

Received on Thursday, 31 October 2002 01:48:00 UTC