Re: Requirements for reliable message delivery

On Tue, 20 Nov 2001 15:17:54 +0100 Brian E Carpenter 
<brian@hursley.ibm.com> wrote:

> I'm wondering if people have had a chance to look at this draft.
> It describes a gap that probably needs filling.

Hmm ...

Not bad from a requirements point of view.   However.

1.   The assertion concerning the applicability of BEEP at the end of the 
document is well ... rubbish.   BEEP as a basic framework is in a much 
better position of providing the feature requirements described.   This is
opinion based on implementation experience with both BEEP based protocols 
and HTTP.   Primary among these is the clear separation of framing and 
control semantics from payload semantics.   BEEP is *designed* for this 
type of semantic loading in the protocol framing itself. 

2.   Work in this area is already proceeding in major ways on at least 
two other fronts -- XML Protocol (W3C) and EBXml (oasis).   This 
problem has been well examined, furiously debated and concensus built 
(more or less) between the two groups.  Not sure why this work needs to be
replicated again.

3.  One of the major outcomes of the discussion (above) is that the 
binding of reliability to a particular set of application protocol 
extensions is not a good idea.   In fact, in a true messaging sense, 
asynchronicity and store and forward situations are a reality that MUST be
dealt with.   Asynchronous models are not dealt with at all.

 ***

I will note that I have now seen, at least on a couple of occaisons, some 
verbage from IBM on the "new reliable messaging standard" that IBM will be
promoting through the IETF.   I presume that this is one and the same.  I 
have to say, that that presumption bothers me a bit.   

Cheers.

---
Steve Hole
Chief Technology Officer -  - Billing and Payment Systems
ACI Worldwide - MessagingDirect
<mailto:Steve.Hole@MessagingDirect.com>
Phone: 780-424-4922

Received on Tuesday, 20 November 2001 16:36:48 UTC