A note on the challenge presented by open content models

As a result of discussions during today's teleconference, and related
work for support of features and the like, it occurs to me that existing
schema models and languages are not terribly well-adapted to open
content models.  I thought that I would share these thoughts (lucky
you!).

At issue is the fact that there is, at least conceptually, a schema for
WSDL.  It has an open content model, which can be represented in W3C XML
Schema as any, namespace ##other.  This is true for almost all elements
in WSDL.

Particular extensions, as a rule, define a sort of micro-vocabulary
intended for use within the context of given extensibility elements in
WSDL.  That is, the elements and attributes defined for a DIME binding
(for instance) would have particular constraints.  Some should appear as
children of wsdl:binding, others as children of wsdl:operation, others
as children of wsdl:input, wsdl:output, or wsdl:fault.  There might also
be elements or attributes intended to decorate wsdl:service or
wsdl:port.

But there is no way, so far as I know, in any schema language, to
express the required *parent* of an extensibility element.  It's an
interesting omission, once one considers it.  A deliberately open
content model, in which the extension specifications constrain
themselves (in effect) is not contemplated, and not supported
(admittedly, formalizing such a thing carries some interesting security
issues).

This is not an issue for resolution, just for its thought-provoking
qualities.  It appears that the path being traveled by WSDL (an open
content model for the base schema, with extensions expected to specify
their preferred or required inclusion locations) is not well known.  It
may present obstacles in terms of defining schemas for such extensions.

Amy!
-- 
Amelia A. Lewis
Architect, TIBCO/Extensibility, Inc.
alewis@tibco.com

Received on Thursday, 17 October 2002 13:33:53 UTC