RE: concatenating web services

> From: Christopher B Ferris [mailto:chrisfer@us.ibm.com]
> Sent: Monday, February 10, 2003 6:44 PM
> To: Don Box
> Cc: xml-dist-app@w3.org
> 
> 
> Don Box wrote on 02/10/2003 12:11:35 PM:
> 
> <snip/>
> >
> > For what it's worth, the team I work on (we build the SOAP stack for
our
> > company's operating system) has done a fair amount of navel
> > contemplation on this one. Our primary conclusion was that because
SOAP
> > has no notion of message identity, intermediaries have a great deal
of
> > freedom. That stated, here are some guidelines to think about:
> >
> > 1) Intermediaries SHOULD NOT contradict the intention of the
original
> > sender and the ultimate receiver. When these two conflict, the
ultimate
> > receiver's intention wins.
> 
> Intention for what? This guideline seems to me to have a very
RPC-style
> bias. In a
> pub/sub environment, neither the intent of the message publisher nor
that
> of
> the message's subscriber(s) is necessarily known to any other than
itself.
> In fact,
> the publisher may not be aware of the existance of the subscriber(s).
How
> is an
> intermediary to "know", or even guess at the intent of the ultimate
> recipient?
> Can the intent of the original sender be "known" to be anything more
than
> that
> it intends the message to be delivered to the ultimate recipient?

Sure. For example, based on the namespace affiliation of the root
element of the SOAP message, I am indicating that I intend all nodes
along the message path to adhere to a specific version of SOAP.

In the case of broadly-adopted SOAP modules/header blocks, the presence
of a WS-Security header block indicates that I intended you to know that
this message was indeed from me (for some definition of "me"). 

Or it may be as simple as "I want you to know that the third element
contains a double (or at least that was what I intended when I formed
the message)." 


> >
> > 2) Intermediaries SHOULD NOT change the "action" of the message. If
you
> > are going this far, consider becoming an ultimate receiver.
> 
> What is the "action" of the message?

Ahhh, sorry for the old-school interpretation. What I meant to say was
"the set of QNames used as GIs for SOAP header blocks + the ordered
sequence of QNames used as GIs for soap body parts" :-)

DB

Received on Monday, 10 February 2003 22:35:45 UTC