Re: New article published: xml:lang in XML document schemas

It seems to me that we are discussing s.t. which is not the topic of the  
article. The article is a FAQ about whether you sould use xml:lang or s.t.  
else. The topic of our discussion is rather a side issue for this article,  
and for that reason I don't see a harm in leaving the article as it is.  
Our issue seems to be a QA "Do I need to declare xml:lang in my XML  
document schema?", with the aspect of different schema languages and  
another aspect of testing constraints of RFC 3066 or further constraining  
them. The discussion about this was very helpful and could be an input to  
such a short FAQ. Any thoughts on this?

Regards, Felix.

On Fri, 25 Nov 2005 19:08:48 +0900, John.Cowan <>  

> Sebastian Rahtz scripsit:
>> It seems to me that you can't have it both ways. Either xml:lang is a
>> universal
>> attribute which is tied to RFC 3066 (or its sucessors), or its not. What
>> if your schema
>> defined xml:lang to accept only integers? How do you manage the
>> contradiction?
> The XML Recommendation gives the purpose of xml:lang, but does not make  
> it
> normative.  Applications of XML are free to make use of xml:lang in any
> way desired; interoperability is served by not allowing the value to be
> other than an RFC 3066 identifier or the empty string, but there is  
> nothing
> in the XML Recommendation to prevent values from being more tightly  
> constrained.
> This could be achieved in XML Schema by importing the XML namespace  
> using an
> xs:redefine element and narrowing the definition of xml:lang.
>> And what about xml:id? are we allowed to redefine the datatype of that??
> Not in a DTD, at least.  The published schema for the xml: namespace also
> forces xml:id to be of type xs:ID.  Authors of RELAX NG schemas may do  
> what
> they please, but are urged not to.

Received on Friday, 25 November 2005 19:02:22 UTC