RE: [ws-rx] i010 - a proposal

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

Jacques,
  Sorry for the delay - I like
the direction you'd headed with this.   Hopefully, I'll be able to
write up a more crisp list of changes this week.  Included in there
I think will be the removal of the text that says the CreateSequence's
wsa:To is the RMD and that the wsa:ReplyTo is the RMS.  It was added
to the spec late in the game and I don't think it adds anything  -
in fact hurts by encouraging the idea of a single endpoint for the RMS
and RMD.  
thanks,
-Doug

Jacques Durand <JDurand@us.fujitsu.com>

09/23/2005 06:56 PM

To	Doug Davis/Raleigh/IBM@IBMUS, ws-rx@lists.oasis-open.org
cc
Subject	RE: [ws-rx] i010 - a proposal

Doug:
 
I believe there is a way to
achieve what you are trying to do without resorting to extending the spec
to handle multiple RMS or RMD, which I strongly advise against (so many
statements in WS-RM assume that sequences happen over a single RMD and
a single  RMS - departing from this would mean a tedious "model"
upgrade I believe.) 
 
I know it may look like just
a modeling game, but I'd rather :
- make the definitions of RMS
and RMD abstract enough so that they tolerate distributed implementations,
Rather than:
- introduce the concept of
 a sequence spanning "multiple RMSs / RMDs".
After all, if I understand
your point at the f2f, you still expect these distributed RMDs to behave
virtually like a single one.
 
In order to do this - and also
to allow for spanning several endpoints, in the sense of Service endpoints
- we could relax the association RMD / endpoint. For example, the core
definitions could be:
 
Line 182: - RMS: An addressable
SOAP processing entity able to execute all functions associated with sending
reliable messages, as specified
in this document (as well as in WS-RM Policy). 
 
Line 183: - RMD: An addressable
SOAP processing entity able to execute all functions associated with receiving
reliable messages, as specified
in this document (as well as in WS-RM Policy).. 
 
Line 80:  I would suggest
simply: "It defines a messaging
protocol to identify, track, and manage the
reliable delivery of messages transmitted
between a source party and a
destination party." (avoiding
to mention any number of parties)
 
There also several places where
"endpoint" is used abusively instead of RMD or RMS.
In Section 2: (L-147, L-150...)
use the term "RMD" instead of just "endpoint".
More generally, endpoint should
mean either service endpoint or client endpoint (e.g. rename " AD
endpoint",  "AS endpoint"), which may have a different
address than the RMD or RMS.
 
In Section 3.2 and elsewhere,
expressions like "RM Source endpoint."
should be reduced to just "RM Source", given that an RMS or RMD
might not be addressable as the same endpoint as the Source or Destination.
 
Jacques

---

From: Doug Davis [mailto:dug@us.ibm.com]
Sent: Thursday, September 22, 2005 6:46 AM
To: ws-rx@lists.oasis-open.org
Subject: [ws-rx] i010 - a proposal
 

(oops - I meant to hit send a couple of days ago)

I took an AI to write up a formal proposal for issue 10:

Issue 10, as stated, deals with whether or not a single RM sequence can
span multiple ports.  Looking at this issue some more though I believe
the broader issue is really whether a single RM sequence can span multiple
endpoints as well.  Should the RM spec limit a single RM sequence
to just a single RM Source and RM Destination or should it leave it up
to the implementation to decide?  I believe it should leave it up
to the various implementations.  If an implmentation is smart enough
to share RM state across multiple endpoints then there is no reason why
a single sequence could not be used to deliver messages to all of them
(and to be clear I do see multiple RMSs and RMDs being an option).  So,
with that mind I'd like to propose the following changes to the specs:

Using this version of the RM spec: http://www.oasis-open.org/apps/org/workgroup/ws-rx/download.php/14548/wsrm-1.1-spec-wd-03.pdf
Line 80: 
        remove the word "exactly" in the
      sentence "...delivery of messages between exactly
two parties,..." 

To the end of line 439 add:
        This specification makes no restriction of
the plurality of the 
      RM Source or RM Destination.  If implementations
can support 
      a single RM Sequence spanning multiple WSDL ports
or even 
      multiple service endpoints then they are free to do
so.  However, 
      it is out of scope for this specification to define
how this 
      ability is communicated or achieved.