Re: [xml-dev] Backward and forward compatible schemas ... Relax NG -->

[noah_mendelsohn@us.ibm.com]

Roger Costello writes:

> However, the new schema is not legal; an optional element prior to an
> <any/> is not legal, as it makes the content of Book
> "non-deterministic";

True in W3C XML Schema 1.0;  not true in the proposed Schema 1.1.  For 
exactly the reason you describe, I.e. that to make languages extensible 
one wants to write schemas just like this, Schema 1.1 will change the 
Unique Particle Attribution constraint, which is the one that enforces the 
determinism in question here.  In schema 1.1, UPA will be violated only if 
the same content in some instance could match two or more element 
particles or two or more wildcard (<xsd:any>) ;  particles.  A wildcard 
"competing" witn an element is OK in 1.1.

> Do you agree with my assertion: Relax NG supports backward and 
> forward compatibility 

I'll take your word for that.  Haven't used RNG in enough real scenarios 
to be sure.


> but XML Schemas does not?

Agree for Schema 1.0, disagree for Schema 1.1.

Noah

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