Re: [ws-rx] i089 - a revisted proposal

[Doug Davis <dug@us.ibm.com>]

DougB,
Answering your notes slightly out of
order...
- the latest proposal is the one pointed
to by proposal #4:  http://lists.oasis-open.org/archives/ws-rx/200604/msg00129.html
- our proposal allows for the server
to still use the anon backchannel if it is still available - that is no
different than normal anon replyTo.  What is different is when the
backchannel is not available.  This can happen under two cases - one,
the socket is closed prematurely or two, RM is being used to deliver the
response and therefore unacked resends must use a backchannel (since the
replyTo was the polling URI).  So, there should be no additional performance
hit in cases where the network connection stays up.  It case it wasn't
clear, the polling URI is meant to match the semantics of anon when things
go well.

-Doug

Doug Bunting <Doug.Bunting@Sun.COM>
Sent by: Doug.Bunting@Sun.COM

05/01/2006 03:14 PM

To	Doug Davis/Raleigh/IBM@IBMUS
cc	ws-rx@lists.oasis-open.org
Subject	Re: [ws-rx] i089 - a revisted proposal

Doug,

I have two questions on this most recent proposal:

    * Since the threads continue on your earlier proposal, I'm
wondering
      which one you consider current?  That is, did
this one take over
      though many of us are responding to earlier versions?
    * I don't understand how this proposal addresses one of the
      important issues I have heard discussed in this TC.
 Whether or
      not the problem is addressed in this TC, there might
be an issue
      with GetQuoteResponse (to use your example) content
that was
      expected on the anonymous URI back-channel but was
not available. 
      The original GetQuote may be acknowledged without
completing the
      higher-level GetQuote / GetQuoteResponse MEP.  If
I understand
      your proposal correctly, you are suggesting the GetQuoteResponse
      would be used instead of the anonymous URI back-channel,
not after
      a specific back-channel is no longer available?  Is
that correct? 
      If so, it seems very inefficient because it requires
an additional
      Request / Response exchange for every higher-level
response,
      including those responses almost-always available
in time for the
      anonymous URI back-channel.

thanx,
    doug

On 27/04/06 07:49, Doug Davis wrote:
> Based on feedback we've received I've attached an updated proposal
for 
> i089.  The basic idea is still the same but I think we've cleaned
> things up quite a bit and eliminated some of the confusion that some
> people thought the old proposal introduced.  This one is pretty
small 
> and still addresses all of the use-cases we've heard about.  The
> biggest change is that we've made it more clear that GetMessage is
> designed to simply (re-)establish a transport-specific back-channel,
> nothing more.
> (sorry, no cute poem :-)
> thanks,
> -Doug