RE: section 2.2.22 Message Exchange Pattern (MEP)

I've been trying to understand this thread, and it seems to me that
there are at least two underlying approaches that are considerably
different.

One frame of mind seems to say that there are message patterns that are
MEP's and there are message patterns that are not and must be considered
to be choreographies, which are more general.  That is, MEP's are a
strict subset of choreographic message patterns and it is clear and
important that some things are MEP's and others are not. 

If that this the case, there seem to be two litmus tests for determining
if a message pattern is an MEP:  1)Whether the message pattern has
previously been considered an MEP by another group (do only W3C groups
count?); 2)Whether the message pattern satisfies a logical criterion
(that I think has been manufactured pretty much with satisfying 1) in
mind).

The other frame of mind seems to be that an MEP is a message pattern,
period, let the chips fall where they may.  No distinction between MEP's
and choreographic message patterns.  The motivation here seems to be
that in practice it may be very difficult to come up with a logical
distinction that will cleanly fence off the things that have previously
been called MEP's, without (logically) allowing a bunch of other things
that are very different from the thingies previously called MEP's or not
including some thingies previously said to be MEP's.

These approaches are, I believe, different.  That is, it appears to me
that MEP has a fundamentally different meaning in these two cases.  I
think that we must choose between them.

(Ritual combat anyone??)

I personally think it might make sense, and be useful, to define MEP's
in some SIMPLE way that makes them a subset of choreographies -- even if
in the process we end up saying that things called MEP's by other groups
are not what we think of as MEP's.  I guess the question would be just
how many of the former MEP's got thrown out and how important we think
they are.  The alternative of making up a definition that fences off
exactly the right territory seems to me to lead potentially to a
definition that is hard to understand, and I guess I would prefer
opening it up to "everything".

-----Original Message-----
From: David Booth [mailto:dbooth@w3.org] 
Sent: Thursday, July 10, 2003 5:14 PM
To: jones@research.att.com; Christopher B Ferris
Cc: www-ws-arch@w3.org
Subject: Re: section 2.2.22 Message Exchange Pattern (MEP)



At 10:49 AM 7/10/2003 -0400, jones@research.att.com wrote:
>  2.2.22.1 Summary
>  A message exchange pattern is a template for the exchange of messages

> between agents that arise from a message and its responses, if any.

I don't think the definition of MEP should be restricted this way.   The

addition of the phrase "that arise from a message and its responses, if 
any" makes this definition unnecessarily restrictive.  In fact, this 
defintion is not even consistent with either WSDL 1.2 or SOAP 1.2 today!

For example, in WSDL 1.2, the "Multicast Solicit Response"[1] pattern
may 
involve a sequence of THREE messages: (1) the initial "solicit" message;

(2) the normal response message; and (3) a fault message that is
returned 
as a result of the response message.

And in SOAP 1.2, the SOAP 1.2 definition of MEP does not restrict the 
concept of MEPs to only those patterns that "arise from a message and
its 
responses".  In fact, the SOAP 1.2 definition of MEP does not restrict 
either the number of messages or the number of nodes involved.

I think it makes more sense for our WS Architecture to define MEP more 
broadly, and recognize that MEPs may range from simple to complex.  Some

languages, such as WSDL, may only deal with simple MEPs involving only 
sequences of one, two or a few messages or nodes (such as request or 
response).  Others, such as choreography, may permit very complex MEPs
to 
be described (presumably out of simpler building blocks).  Both WSDL and

SOAP define certain, specific MEPs, (and clearly the relationship
between 
them should be clear), but these are only a few of the possible universe
of 
all "Message Exchange Patterns".

I propose simplifying our definition of "Message Exchange Pattern" to:

"2.2.22.1 Summary
  A message exchange pattern is a template for the exchange of messages
  between agents."

1. 
http://dev.w3.org/cvsweb/~checkout~/2002/ws/desc/wsdl12/wsdl12-patterns.
xml#multicast-solicit-response
2. http://www.w3.org/TR/2003/PR-soap12-part1-20030507/#soapmep


-- 
David Booth
W3C Fellow / Hewlett-Packard
Telephone: +1.617.253.1273

Received on Friday, 11 July 2003 16:18:30 UTC