RE: Composition by extension in namespaced schema

Michael Kay writes:

> For some strange reason which I've never understood, local element
> declarations (i.e. <element name="x"/> within a complex type) default to
> being in no namespace, rather than defaulting to the target 
> namespace of the
> schema.

How to treat this was probably the most contentious question in the design 
of Schema 1.0.  It was the so-called issue 208 and caused so much 
disagreement that the number 208 became a standing joke in the group for 
some time.  I agree with your implication that defaulting to qualified, or 
in fact fixing the choice at qualified, would have been sensible.  Others 
felt that (a) attributes were a precedent, not withstanding that the 
Namespaces in XML Recommendation treats them asymmetrivally regarding 
defaulting and (b) there are/were those who felt that using the same 
qualified name with two different local meanings conflicts with the spirit 
of making the names globally unique, and through use of URI-grounded 
QNames, offering the promise that the elements or their names are 
resources on the Web. It's a long discussion that we don't want to repeat 
(trust me!), but the "strange reasons" were along those lines.  Since the 
group went in circles on this and adherents to both preferences felt 
strongly, the switch was introduced. 

Noah

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