Re: Versioning of XML Schema and namespaces

Another point may be pertinent to this discussion:  schemaLocations on 
imports as well as xsi:schemaLocations are indeed hints insofar as the 
core Schema Recommendation is concerned.  You can write conforming 
processors that do or don't honor them, either unconditionally or 
depending on the circumstance.  What may be less clear is that nothing 
prevents anyone from writing a specification for some sort of processor 
profile.  For example, I could write my own specification and say that to 
be a "Noah-compatible" schema processor you must (a) conform to the schema 
recommendation and (b) honor all schemaLocation hints, reporting a fatal 
error if any fail to resolve to a schema document for the designated 
namespace.

Furthermore, there is ongoing consideration to having the schema workgroup 
define some such profiles for Schema 1.1.  Most of the emphasis in the 
proposed profiles is on subsetting the reporting of the PSVI, but there is 
also consideration of whether schemaLocations are honored.  Of course, all 
this only gets interesting to users as implementors of processors choose 
to support such profiles. 

In any case, you are today free to write or use processors that honor 
schemaLocation hints in whatever manner meets the needs of your 
applications.  You or anyone can promote specifications for such 
processors to increase interoperability among multiple implementations. 
Maybe or maybe not the Schema WG will at some point help you by calling 
out profiles for schemaLocations in particular.

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