Transport-specific SOAP semantics - was Re: Visibility

Mark Baker wrote:
> On Tue, Feb 25, 2003 at 09:09:31PM -0500, Champion, Mike wrote:

>>Whoa yourself :-)  If a SOAP message has a semantic difference if it is
>>transported (I use the term deliberately) over SMTP than if it is POSTed
>>over HTTP (or send on a floppy disk by sneakernet), then some deep
>>requirement of WSA (and SOAP 1.2, IIRC) is not being met. 
> 
> 
> Do you recall the recent xml-dist-app discussion about the semantics of
> a SOAP message including the method of the underlying protocol?  Noah
> wrote;
> 
>   'I think it's useful and appropriate to separate the term "message"
>   from "envelope".   I think the destination and web method are surely
>   part of the message.'
>  -- http://lists.w3.org/Archives/Public/xml-dist-app/2003Feb/0005
> 
> Which says to me that sending a SOAP envelope with HTTP PUT means
> something different than sending it with POST (and any other
> application protocol method, for that matter).

If "means something different" affects the semantics of the SOAP
exchange at the application level, I think you've just opened
a large can of trout. Suppose I want to write a web service which
can support client interactions over HTTP, BXXP, JMS, or RFC1149
avian transport. SOAP over JMS (or SOAP over carrier pigeon) doesn't
have any notion of PUT or POST. And even if you view this as something
conditional (i.e. "if this SOAP message arrived via HTTP then if PUT
then foo() else bar()"), consider a SOAP gateway which receives a
message via JMS and dispatches it to the service using HTTP. Should
it use PUT or POST? Etcetera.

Whatever happened to abstraction as a design principle?

Geoff

Received on Wednesday, 26 February 2003 08:31:36 UTC