- From: Jeremy Carroll <jjc@hpl.hp.com>
- Date: Sat, 26 Oct 2002 06:19:42 +0200
- To: w3c-rdfcore-wg@w3.org
In the after hours discussion on 25th Oct, DanC indicated a preference for "An untyped literal is a string optionally combined with a language identifier" to my "... a string combined with a (possibly empty) language identifier". His rationale is desiring semantic interoperability between xsd:string and the untyped literals. i.e. <rdf:Description> <eg:prop>foo</eg:prop> </rdf:Description> entails, and is entailed by <rdf:Description> <eg:prop rdf:datatype="&xsd;string">foo</eg:prop> </rdf:Description> I have rejected this proposal, subject to WG review, for the following reasons. + <a href="http://www.w3.org/XML/xml-V10-2e-errata#E41">XML erratum 41</a> permits xml:lang="" and clearly indicates this is equivalent to not having one. Hence at some point we need to merge these into a single case. I find it more elegant to have that single case being "" rather than missing, for reasons of uniformity. + When doing language specific processing on langauge tags the general algorithm indicated by RFC 3066 is that one should try and match prefixes - e.g. if looking for en-US and you find en-AU then that is a better match than finding fr. The empty string fits with this algorithm, whereas optional does not. + I think DanC's proposal makes the impact on the meaning of adding an xml:lang to the rdf:RDF element, or to an enclosing XML document (not an uncommon practice) too severe. I suspect we may get bug reports along the lines of - this RDF works here, and it doesn't work here - as people copy paste RDF between different enclosing XML. Jeremy
Received on Saturday, 26 October 2002 00:21:21 UTC