RE: SV: What's a valid instance...James Clark

> >In addition to the rationales you mention, I'd also mention that 
> >specifying a required root tends to make reuse of schema modules in 
> >other contexts harder.
> 
> Surely that just means that the specification has to take 
> that into account?
> Some wording about how the required root is determined via 
> import order and so forth?

You missed the point. In the real world you usually have a family of
document or message types that share components in common (often hundreds of
them). There are many problems in managing such families of messages, and
XML Schema doesn't solve all of them, but at least it recognizes the
problem. 

I'm not saying it would be bad to have an optional clause in the schema that
says whether or not a given element declaration is allowed to serve as a
validation root; but it's certainly highly desirable to allow a schema to
define lots of possible validation roots.

Michael Kay
http://www.saxonica.com/