RE: RM Group Definitions

[Nicholas Kassem <Nick.Kassem@eng.sun.com>]

At 02:01 PM 8/30/2000 -0400, mwsachs@us.ibm.com wrote:
>Normally, when a party needs to know whether a message was received before
>sending another, they use a request-response protocol at the business
>application level.  Those are the requests and responses laid out in the
>business protocol section of the tpaML proposal.
>
>Using the RM protocol for this kind of flow control is definitely mixing
>the functions of the MS and application levels.

You are absolutely right Marty. One notion that helps keep me on an even 
keel (some would argue I capsized a long time ago) is to think only in 
terms of asynchronous messaging at the BP level. Period. Now at the MS 
level the transport may be synchronous or asynchronous (that's a 
private  implementation/configuration detail) but the programming model at 
the BP level is always the same. IMO, maintaining a consistent programming 
model will be key to adoption of  ebXML TR&P. If we tie the programming 
model to the messaging choreography we may well be setting ourselves up to 
fail. It's entirely practical to have a request-response application 
protocol that enables an application  to work through it's business logic 
in a consistent and well defined way. So, as for the notion of "grouping", 
I have yet to see a convincing use-case (without significant hand-waving) 
that justifies the complexity. I would very much appreciate it if someone 
could post such a use-case, making a clear distinction between the MS view 
and the BP view. Thanks in advance.

Regards,
Nick