Re: WSDL happiness

Couple of qualifications:

First Everything Is Conditioned by our desire not to go through Yet 
Another Last Call on all the WSDL specs. This would be a huge disaster.

On Sep 28, 2005, at 4:27 PM, Kendall Clark wrote:

> Folks,
>
> I sat in on the WS-Desc F2F via telephone yesterday while they 
> discussed the
> three issues raised by DAWG re: WSDL 2.0:
>
> 1. outputSerialization limitations
> 2. faultSerialization limitations
> 3. POSTing application/x-www-urlencoded
>
> The WS-Desc members decided to rework how they handle input, output, 
> and
> fault serializations, choosing a design based roughly on HTTP Accept:
> header, though I'm not totally sure of the details because the 
> connection
> was very bad. At any rate, with the anticipated changes to WSDL 2.0, 
> we'll
> be able to more accurately describe our protocol as a web service. 
> Which is
> a good thing.

The risk here is that the solution may be too complex to just pop in 
without another last call. We have two strategies to avoid (some) of 
this:

	1) As we see from 3, the HTTP binding is just buggy and underreviewed. 
So, the likely current plan is to split it out and let it do a separate 
LC again. We hope that can be short and sweet and we catch up with the 
rest of specs in CR.
	One thing should be clear: The HTTP binding is *not* designed to be a 
complete description language for everything you can do with HTTP. HTTP 
is a complex protocol, esp. in the wild, and it would be a major effort 
to do so. What the WG tried to do, esp. giving the overwhelming 
interest in group and among vendors for the SOAP binding over the HTTP 
binding, is cover descriptions of a useful and interesting set of HTTP 
services. If there is interest in more coverage, we would hope that a 
future working group would build on community based extensions.

> Re: (3), there was some question, because the WG apparently had 
> intended a
> design whereby it was possible to POST urlencoded In Messages to an
> endpoint, but after chasing down all of their specs, which was not an 
> easy
> task, I'd concluded that it wasn't possible.

(Rightly; in fact, more so than he thought!)

> After several conversations with Hugo, the W3C contact for WS-Desc, he
> encouraged me to file a LC about (3). When I spoke to the WG 
> yesterday, they
> assured me (3) was w/in their design and I took an action to figure 
> out what
> parts of the spec had misled me.
>
> Apparently after I was no longer able to call in, they started 
> reviewing
> some of their specs with (3) in mind and they concluded that I'd been
> correct, that the language as-written made (3) illegal.

There was a design bug and a spec bug and a something in between. We 
are rectifyign that.

> While this will require, or so it would seem, some spec surgery on 
> their
> part, our preferred protocol design will be able to be expressed with 
> WSDL
> 2.0, eventually, after they make the requisite changes.
>
> For my money, this was a good example of WGs working together 
> efficiently
> and courteously.

Yes. I do not speak officially for the group, but this was good and 
useful feedback that helped us realize some serious problems before CR. 
Thanks!

Cheers,
Bijan.

Received on Thursday, 29 September 2005 00:37:37 UTC