Re: SOAP intermediary - issue 70 (cont'd)

+1

There are going to be cases where there are network nodes which do
not perform SOAP processing, do perform SOAP processing, or perform
SOAP processing dependent upon some external criteria (HTTP headers,
URI, phase of the moon, etc.). Also, SOAP is intertwined with
underlying protocol features, esp. routing, leading to general
confusion about what a SOAP node is and is not.

People composing applications with SOAP are going to have to consider
these things, just as people who use HTTP over IP have to think about
getting an IP address. 

From our viewpoint, none of this matters; we can only talk about the
range of "SOAP processors", not "the network node that could be a
SOAP processor". However, it's important for us to have a rigid
definition of what a SOAP node is, and how it is identified. I think
that definition blesses any thing which follows the processing model,
step-by-step as a SOAP node.



On Tue, Oct 16, 2001 at 09:16:39AM -0400, Noah_Mendelsohn@lotus.com wrote:
> Jean-Jacques Moreau writes:
> 
> >> BTW, I do not think "forwarding" is a direct consequence of
> >> "processing". The intermediary may decide not to process the
> >> message, and still do forwarding.
> 
> I don't think so.  The spec says you must assume the role of "next"
> if you are a node, including an intermediary node.  If there is a
> mustUnderstand to next you MUST process it.  Even doing those
> checks is "processing" in the terminology of chapter 2.  So, I
> think every SOAP node MUST process each message.  It is not
> necessarily required to assume any roles, other than next, that
> would result in its actually processing any particular headers, but
> it must at least follow the chapter 2 rules.
> 
> One can imagine software that would do other useful things with
> SOAP messages, but such behavior is not covered by the SOAP spec, I
> think, just as HTTP doesn't explicitly say that you can buffer
> messages on disk if you like.  I don't think we should get into the
> business of nodes that don't process messages.  IMO, every node,
> including intermediaries, should follow the rules in chapter 2,
> which is what it means to "process" a message.

-- 
Mark Nottingham
http://www.mnot.net/

Received on Tuesday, 16 October 2001 21:41:14 UTC