Re: WSDL 2.0 element order enforcement in schema

Hi, Roberto,

On Fri, 15 Jul 2005 17:41:37 -0700
roberto chinnici <Roberto.Chinnici@Sun.COM> wrote:
> Amelia A Lewis wrote:
> > Hmmm.  I'd recommend that we document this in our schema, including
> > text to the effect:
> > 
> > The content model of the documentation element as represented in schema
> > is not complete.  Validators MUST implement the constraints as
> > specified in [specref], which permits extension elements in more
> > locations than we can show in a schema that abides by the UPA
> > constraint.
> I've got an action to implement this and I find the text confusing.

Yeah, me too.  :-)

> Did you really mean the *wsdl:documentation* element or was it the
> *wsdl:description* element whose content model is not complete?

wsdl:description.  I had a braino while writing it.

> Furthermore, is it true that the spec permits extension elements in
> more locations than the schema does? Perhaps it depends on what
> you mean by "location". It seems to me that being schema-valid is
> necessary but not sufficient in order for a document to be spec-valid.
> It follows that the spec cannot "permit more" than the schema does,
> doesn't it? Maybe by "more locations" you meant that in 2.1.2 [1]
> the buckets identified by the last bullet of (2) and the last bullet
> of (4) respectively are indistinguishable when the wsdl:types element
> is not present?

Hmm.  I will have to look at the current version of the schema in order
to answer this.  But, in general, I think the answer is that extension
elements may appear in more places than the schema indicates, because
typically adding another wildcard triggers UPA.

> One last point. In the proposed text I see a MUST requirement on
> something called a "validator". Isn't that tantamount to reintroducing
> the concept of "WSDL processor"?

Argh.  Probably.  Is there a better way to describe the problem?

Our schema is incomplete.  Schema validation alone is insufficient.
This is a longstanding, known problem, related to UPA and wildcards.
There are increasingly complex workarounds; the tradeoff for increasing
complexity is decreasing comprehensibility.  The WG decided not to go
with the substitution group solution, which is closest, instead using
the model we currently have, which is not ideal.  I'll go look again, I
guess, and see if I can't propose better text.  But I think the above
suggestion of "more locations" is, in fact, correct.  Same problem with
current version of WSDL 1.1 schema, last I checked.

Amelia A. Lewis
Senior Architect
TIBCO/Extensibility, Inc.

Received on Monday, 18 July 2005 14:27:20 UTC