Re: On RDF/I18N issue

Jeremy Carroll wrote:

> Guus (offlist):
> >Jeremy, any progress on this action?
>
> >ACTION: Jeremy to send his discussion of some of the issues re
> >xml:lang and literals to WG.
>
> Original message archived at
> http://www.w3.org/mid/3F42466F.80906@hplb.hpl.hp.com
>
> see html attachment (also on this message).
>
> The bit missing is the explanation of the issue which I will summarize 
> now:
>
>
> consider the xml:lang attribute in this
>
> <rdf:RDF xml:lang="fr">
>   <rdf:Description rd:ID="x">
>     <eg:prop rdf:parseType="Literal">
>        <foo>chat</foo>
>     </eg:prop>
>   </rdf:Description>
> </rdf:RDF>
>
>
> According to XML, and the RDF last call documents, the xml:lang is in 
> scope  for the string "chat", and it is understood as French.

Nope. According to XML 1.0 the xml:lang attribute has no required 
semantics. From the "Annotated XML 1.0 Specification by Tim Bray:
[[


      xml:lang Has No Required Effect

Like the xml:space attribute and the Standalone Document Declaration, 
the xml:lang attribute doesn't have any effect on the required behavior 
of XML processors or applications. It's just a pre-cooked way we provide 
for authors to provide this information to any downstream applications 
that might care.
]]

So as I read it, the current RDF docs defer to exclusive 
canonicalization which apparently doesn't consider xml:lang scope. Fine. 
I see no _requirement_ per XML 1.0 that it so be considered (on the 
other hand neither does XML 1.0 _prevent_ external xml:lang's from being 
considered in scope -- XML 1.0 doesn't ever say much about semantics, 
its an almost entirely syntactic specification by design.)

There's also nothing preventing you from writing:

<rdf:RDF>
  <rdf:Description rd:ID="x">
    <eg:prop rdf:parseType="Literal">
       <foo xml:lang="fr">chat</foo>
    </eg:prop>
  </rdf:Description>
</rdf:RDF>

so I don't see the big deal here ... sure it may require a number of 
extra xml:lang's but the RDF/XML syntax is, err, not the most compact 
for a variety of other reasons. I hardly think this is the worst issue 
with RDF/XML, and feel that way too much time has been devoted to the 
semantics of xml:lang -- sort of like trying to wring water out of a rock.

Jonathan

Received on Monday, 1 September 2003 16:25:57 UTC