Processing instruction to designate a schema as subset of another?

[Stephen Green <stephengreenubl@gmail.com>]

Would it be feasible to use a processing instruction at the
top of an xml schema to designate it as a subset of another
schema (in the sense Rick Jelliffe recently mentioned on this
list of all instances valid by the subset schema being also
valid by the superset schema)? How would one go about
defining such a processing instruction? Would it have to be
something like DSDL - via ISO? I noticed Rick's mention of
the need for specifying a schema and/or namespace as a
subset of another schema and/or namespace and thought
maybe the problem could get critical mass of interest sufficient
to solving it. I have found the problem exists not only when
the subset schema has a different namespace; it was tricky
to find a way to define a subset without changing the
namespace when first working on subsets for the Universal
Business Language and I think people eventually just wrote
a second schema with the same namespace but without any
way other than an accompanying prose spec to show it as
defining a subset of a superset (the Standard UBL schemas).

Maybe to solve this a processing instruction could be defined
which specifies within a subset schema that it is a subset of
schema at location ABC, either same namespace or
namespace XYZ. There might also need to be a way to tell
tools (if feasible, I've no idea) that they should either allow
nodes of just the superset or not when the subset schema is
given as the schema for the instance. How then to get such
a PI or set of PIs specified and supported?

Otherwise it might be that one has to stick with prose and/or
accompanying (test) assertions such as those of Schematron
or (work-in-progress) OASIS TAG TC's Test Assertion Markup
Language (shameless plug).
---
Stephen D Green