Re: [xml-dev] relax ng and wsdl

On Mon, Oct 17, 2005 at 12:56:36PM +0200, Robin Berjon wrote:
>De Heer, Robert S - San Mateo, CA - Contractor wrote:
>>I have been working on importing a relax ng schema into a wsdl and have
>>run into some problems with namespaces/getting wsdl2java to recognize
>>the relax ng schema. Has anyone done this successfully? Could you
>>provide an example? 
>
>Unfortunately, I don't think anyone has actually implemented RNG support 
>in a WSDL processor (I'd be happy to learn otherwise though). There's a 
>discussion of how it would be done here:
>
>  http://www.w3.org/TR/2005/NOTE-wsdl20-altschemalangs-20050817/

The note is, please be aware, for WSDL 2.0, not WSDL 1.1.  I believe that
wsdl2java, the tool mentioned, operates on WSDL 1.1 documents.

While it is possible that the techniques mentioned in the note could be
implemented for WSDL 1.1, it would be a more complex effort, I suspect. 
wsdl:definitions/wsdl:types *is* an extension point, so it's feasible there. 
Unfortunately, wsdl:definitions/wsdl:message (and child elements, particular
wsdl:part) are not extensible (http://www.w3.org/TR/wsdl#A3), so adapting a
different "type" system (schema language) to WSDL 1.1 would involve redefining
the use of wsdl:part's @type or @element attributes.  The problem extends to
the definition of the SOAP binding schema as well; either the behavior of
certain elements and attributes must be redefined, or alternate constructs
must be defined anew.

This is not to say that it can't be done, just to say that it probably can't
be done as gracefully (by comparison only, mind) as the referenced WSDL 2.0
note does for RNG and DTD.  And that is, in large part, because the WSD WG
made strong efforts to ensure that WSDL 2.0 wasn't tied to W3C XML Schema as
the only possible schema language (and produced the note more as "proof of
concept" than anything else: "here's how you could do some alternative
schema language, go ye and do likewise").

A particular problem when an alternative schema language is defined (for
either version of WSDL) is: what happens when there are competing
definitions for the same construct?  That is, what if the extension provides
an alternate mechanism for pointing at schema constructs, and both pointing
mechanisms exist within some definition?  (for example, what if a message
structure has a pointer both to a DTD definition and to a W3C schema
definition?  These are almost certainly incompatible to a degree, since DTD
definitions cannot "type" elements)  See
http://www.w3.org/TR/wsdl20/#other-types, especially the warning at the end
for the resolution in WSDL 2.0: "behavior is undefined" if there are
definitions in two different schema languages.

So.  WSDL 2.0 is not (yet) in CR, so I doubt that available tools (like
wsdl2java) can cope with its (changed) syntax (it *is* 2.0, not 1.2; there
are *lots* of changes).  Adapting the techniques suggested in the referenced
Note to WSDL 1.1 is likely to be painful at best.  Feedback on the utility
of the extension mechanism for WSDL 2.0 would probably be appreciated,
though.

Amy!
(speaking entirely for herself (and *all* the *little people* (no, sorry,
got carried away there, it's just us (you mean "just me" (oh, right,
thanks)))))
-- 
Amelia A. Lewis                    amyzing {at} talsever.com
The law, in its majestic equality, forbids the rich as well as the poor
to sleep under bridges, to beg in the streets, and to steal bread.
                -- Anatole France, "Le Lys Rouge"