RE: [xml-dev] xmlns empty string

[noah_mendelsohn@us.ibm.com]

Michael Kay writes:

> This is one of those horrible cases where the spec explicitly says that
> xs:include operates on schema components, and then says that 
> its treatment
> of the schema components depends on something ("whose code was 
> qualified")
> that (as far as I can see) isn't a property of the schema component, but
> part of the syntax of the schema document from which the component was
> derived.

Agreed. 

> (What is "code" anyway? I suspect it's a misprint for "form".)

I noticed that too.  Not sure what was intended (though I worked on some 
of the general framework for composition, the particular details of 
include and chameleon processing were not areas I was deep into as the 
spec was being drafted).

> In fact I think this is a red herring. I think the chameleon 
> include affects
> all nested element declarations whose {target namespace} is .absent.,
> regardless of the reason why it is .absent..

Should be that way, I think, but I noticed the ambiguity as I was 
responding to your earlier note.  I openned a bug in bugzilla suggesting 
it work that way.  I think one also needs to be careful about the case 
where a user includes from multiple schema documents with multiple 
targetNamespaces the head document in a chain of chameleons.  I'm guessing 
you want multiple sets of components, one for each of the 
targetNamespaces. Whether users should ever try such a thing is a 
different question.  I have my doubts that chameleon should ever have been 
provided in the first place, but given that it's there, we need good 
symmetric rules for how it can be used.

Wondering whether any further discussion of this should be moved to the 
schema-dev list?  I think we're well over the line where an ordinary 
mortal interested in XML would be getting a bad headache.

--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------