It’s just data

WSDL 1.2

Simon Fell: The WSDL folks are looking to drop use="encoded". As much as the Microsoft folks would like for encoded SOAP to not exist, it does, and a WSDL spec that can only describe a subset of SOAP messages just seems like a recipe for YADL to appear.

As of yet, I've not been involved in WSDL 1.2, but I am curious as to what you see that I don't? With concrete proposals to handle arrays and arbitrary directed graphs, it seems to me that SOAP encoding is a subset of what can be described via document literal.


What about existing - soon to be legacy - systems? Like ... .NET Remoting ;-)

Posted by Ingo Rammer at

As is often the case, I agree with Sam.

As for Ingo's statements about .NET remoting, its our job to get doc/literal support for the entire platform. We've already signed up for this as part of the WS-I Basic Profile, so WSDL/1.2's move in the same direction is a non-issue.

All of this stated, there's nothing stopping a bright young man who's written a book in this area from writing the reflection code to emit such a WSDL document today ... :-)

DB

Posted by Don Box at

As i see it the proposal drops support for describing soap encoded messages in WSDL.
Regardless of whether or not soap encoding is a good or bad thing, i see the loss of this ability as a bad thing.
Whilst soap encoding is in the soap spec, i'd like to see WSDL cover it. if soap encoding really is that bad, then get the XMLP guys to pull it from the SOAP spec.

Posted by Simon Fell at

Simon, I'm still not getting it. Can you provide an example SOAP message that can not be described in document literal terms?

Posted by Sam Ruby at

I'm getting more confused not less confused as time goes on. :(

How about the SOAPBuilders echoStructArray message. I understand that i could write a schema that describes a particular serialization of this, but is it possible to write a schema that describes all the valid serializations of this ?

without a use="..." attriubte how do i indicate that i require a encodingStyle="...." attribute in the request ?

Posted by Simon Fell at

Conversely, I'm getting less confused. I'm starting to see what you are getting at.

This being said, while the choices the WSDL WG are looking to make are hard ones, I believe that having specifications which more precisely identify particular serializations is benefitional both from an implementation and interoperability point of view.

I still believe that identifying a particular serialization that one will accept says little about what other serializations a particular implementation may also happen to accept.

Posted by Sam Ruby at

But do the interop problems with soap encoding come from the soap encoding model itself, or attempts to describe it in WSDL ?

I'd like the base WS specs to be consistent, not for them to support different subsets. Again, if soap encoding is so evil, shouldn't it be pulled from the SOAP spec as well ?

Agreed, however doesn't that result in a schema that can't be used for validation ?

Posted by Simon Fell at

I believe in validators... heck I even co-wrote one! That doesn't mean that I believe that one shouldn't be tolerant about what one accepts.

I still think that there are too many options for expressing the same thing. Note: at the time I wrote that, of the eight possible ways to describe this API, with Axis three produced one binding and the other five produced another. At least all produced bindings, only two produced anything at all with .Net. On the plus side, .Net did properly report that one (or was it two, I can't remember) was invalid WSDL.

Posted by Sam Ruby at

There's a thread about the dropping of use='rpc' from WSDL 1.2 over on SOAPBuilders. Rich Salz is responsible for my favorite quote in the thread Geez, this is starting to sounds like the Monty Python skit. Why is it so hard to make it clear? I've...

Excerpt from Simon Fell at

Add your comment