RE: T2 SyncReply and ReliableMessagingMethod in QualityOfServiceI nfo

["HUGHES,JIM (HP-Cupertino,ex1)" <jim_hughes@hp.com>]

I hesitated to jump into this conversation, given all the air time we had
last year on this topic! I think the group should schedule a teleconference
to discuss this... 

But, IMO, I think that when a FromParty MSH gets a request for RM
transmission of a message, I think the most obvious target is the ToParty
MSH, not the first MSH in a chain of MSHs leading to the ToParty MSH. I
expect that's what the FromParty's application really wants, and we should
configure for this basic case.

Now, when we have intermediate MSHs that can play the RM game, we need to
figure out if composition functions are needed among node pairs and show how
they relate to the overall RM functionality that was requested. One way to
do this is just to let each intermediate MSH node pair decide to use RM if
they want or not, based on knowledge that the overall transmission is RM;
that is, decouple the overall RM algorithm from intermediate node pair
transmissions... if both the message and the MSH-level ACK get through, then
the problem is solved. If the end-to-end ACK never gets back to the
FromParty MSH, then there was failure, whether it was from a broken cable or
a misbehaving intermediate node pair. This would keep it simple at first
pass, and we can work next on how composition functions might evolve.

I joined the conversation because I was troubled by words that said: if
there is an intermediate node link based on MQseries, then things change and
reliability flags don't need to be passed. This thinking misses one of the
main points of ebXML messaging - it should be usable and consistent
regardless of the path taken. IMO, a properly operating Message Service
should not recognize whether the intermediate links are based on MQseries,
html or carrier pigeons (assuming a binding!) - RM is an expression of state
to the sending application that the message got into the persistent store at
the MSH on the far end of the transmission. That's all - but I think that's
what you mean by quality of service?

jim



> -----Original Message-----
> From: christopher ferris [mailto:chris.ferris@east.sun.com]
> Sent: Tuesday, August 07, 2001 10:17 AM
> To: HUGHES,JIM (HP-Cupertino,ex1)
> Cc: ebxml-msg@lists.oasis-open.org
> Subject: Re: T2 SyncReply and ReliableMessagingMethod in
> QualityOfServiceInfo
> 
> 
> Jim,
> 
> The algorithm doesn't extend beyond the next node 
> from the perspective of any given sending MSH node.
> 
> How then does an intermediary affect the basic algorithm?
> 
> I'm not saying that we don't have work to do to refine/fix
> intermediary handling, but in my mind, it is the quality of
> service, not the protocol that extends from end to end.
> 
> The RM protocol is strictly point-to-point.
> 
> Cheers,
> 
> Chris
> 
>