Re: [xml-dev] Extensibility vs WS-I

[Chris Burdess <dog@bluezoo.org>]

Toby Considine wrote:
> Actual implementations, though, should not rely on extensions and substitution groups. WS-Interoperability is likely to be required—and it forbids the use of these in schemas (as I understand it). When we get past the joy of abstract models interacting with valid extensions to actual applications that either know what to do with an xml artifact or they don’t, one must pick and choose, and send messages that will be understood at the other end.
>  
> Is there a standard approach or set of tools for generating and or validating these restricted or profiled schemas? In a perfect world, an artifact validated from the full schema but matching the profiled business rules would be indistinguishable from another artifact representing the same information validated by the profiled, non-extensible schema.

Sans rancune, WS-Interoperability is really not worth the pixels it's written on. As you've pointed out, either the other end understands the schema or they don't. What is comes down to in the end is actual human beings, since the problem domain is semantics, and machines, wonderful as they are, can only deal with the syntactic side of things. You simply can never expect to send an XML message, no matter how beautifully "self-describing" or layered in the honey of a hundred WS-* completed checklists, and have it be understood at the other end if the recipient is not a human being who understands that schema. The solution to this particular problem is to schedule a real-life meeting (phone conference or whatever) and make sure that the developer of the receiving application knows what messages he's going to be receiving and what they mean.