Re: xml:lang in XML 1.0 (3e) and XML 1.1

On Tue, 2005-08-16 at 14:56, Martin Thomson wrote:
> This issue has been raised previously on this list
> (http://www.w3.org/2002/02/mid/8E9F0028F5955844899380433C60E39902A0D2FC@cscrdemsg001.crdc.kp.org;list=www-xml-schema-comments) and it appears that work stalled waiting for RFC 3066bis to complete.
> 
> I have recently run into this problem and since it has been a year, I
> thought it worthwhile raising the problem again.  I am really only
> concerned at this stage about the first issue raised, that is,
> xml:lang is defined to be valid as the empty string.

Having thought about this, and then reviewed some of the
email threads (it would have gone faster in the other 
order), I believe (speaking for myself) that

  1 The type xsd:language should not allow the empty
    string, because the empty string is not a language
    code as they are defined by RFC 1766, RFC 3066,
    and XML 1.0 (any edition).

  2 If XML 1.0 3e allows the empty string as a value for
    xml:lang, with the meaning "clear the lang value",
    then the correct definition for the xml:lang attribute
    associates it not with type xsd:language but with
    a union of xsd:language and the empty string.

I believe that mail from Robin Berjon and Henry Thompson
suggests they take the same view I do.

That seems to suggest that the only things that 
should be done are

(a) the XML Schema WG should decide whether to issue
a clarification for 1.0 on this topic (e.g. a Note
in the text calling attention to the mismatch and
explaining how to deal with it), and if so, the
corrigendum needs to be drafted and approved.

(b) the XML Core WG should change the schema at
http://www.w3.org/2001/xml.xsd to use a correct type
for xml:lang.

If you believe this is the wrong resolution, please let 
us know; otherwise, we'll probably assume you agree 
with us.

-CMSMcQ