W3C home > Mailing lists > Public > xml-dist-app@w3.org > October 2001

RE: SOAP Binding Framework, HTTP Binding, MEP documents

From: Williams, Stuart <skw@hplb.hpl.hp.com>
Date: Fri, 19 Oct 2001 12:12:36 +0100
Message-ID: <5E13A1874524D411A876006008CD059F1926FC@0-mail-1.hpl.hp.com>
To: "'Glyn Normington'" <glyn_normington@uk.ibm.com>
Cc: xml-dist-app@w3.org
Hi Glyn,

> -----Original Message-----
> From: Glyn Normington [mailto:glyn_normington@uk.ibm.com]
> Sent: 16 October 2001 15:27
> To: xml-dist-app@w3.org
> Subject: Re: SOAP Binding Framework, HTTP Binding, MEP documents
> 
> 
> Some initial comments:
> 
> 1. I cannot find a clear statement in these documents about whether or not
> XMLP will define a common behaviour in respect of at-most-once delivery,
> out-of-order delivery, and message loss.

I guess there are two thing here... the 'contract' between SOAP and the
things (applications) that use SOAP and the 'contract' between SOAP and the
bindings/underlying protocols that SOAP uses. 

The TBTF work is focussed on the latter, although the two are clearly
related. In terms of the way the TBTF framework is cast, the contract
between SOAP and binding will be modular in that a binding will declare that
it supports a particular set of transport meps and features. Delivery QoS is
something that I would expect us to cast as a feature and the semantics of
that feature with respect to its use in conjunction with different transport
meps will need to be well defined.

I know this too is not a clear answer to your question, but I hope it gets
you some way to understanding the thinking that's going on.

> 2. R604 says that the XMLP spec. must consider message paths 
> over multiple transport protocols.
> Does the working group intend to consider transports which exhibit
> re-delivery, out-of-order delivery, and message loss?

I think so... the current HTTP binding SOAP 1.1 says nothing about
management of the underlying HTTP connection so I would contend that we're
already in a world that makes no guarantees about delivery order; equally it
HTTP operations can fail, so we're in an environment that is lossy;
multiple-delivery... not sure if there's a failure mode in HTTP that can do
this eg. authentication failures and redirection may cause re-transmission
of a SOAP message... what are the semantics with respect to the delivery of
earlier copies... that's probably a stretch... I'd think that it should be
as if they had not been sent.

Some folks are considering SMTP and I have certainly expreienced the
occasional mail-loop, message loss and re-sequencing of message.

> 3. The HTTP binding describes a state transition in the 
> requesting SOAP node from 'waiting' to 'requesting' in
> the case of a 3xx redirection response. Should the state transition
> diagram in MEP be updated to reflect this?

Well spotted... I was aware of this one, but didn't fix the diagram.
Partially because the direct handling of 3xx in the description is a bit
tentatative.

> Glyn Normington

Regards

Stuart Williams
Received on Friday, 19 October 2001 07:16:35 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:59:04 GMT