W3C home > Mailing lists > Public > public-ws-desc-meps@w3.org > April 2003

Re: Updated document

From: Amelia A. Lewis <alewis@tibco.com>
Date: Wed, 9 Apr 2003 10:34:28 -0400
To: "David Booth" <dbooth@w3.org>
Cc: public-ws-desc-meps@w3.org
Message-Id: <20030409103428.6fc239a9.alewis@tibco.com>

On Tue, 08 Apr 2003 17:17:38 -0400
David Booth <dbooth@w3.org> wrote:

> 
> I've updated 
> http://dev.w3.org/cvsweb/~checkout~/2002/ws/desc/wsdl12/meps-vs-iops/meps-vs-iops_clean.htm
> 
> :
>          Made IOP and MEP definitions more precise
>          Made assumptions more explicit

I disagree with at least one of these assumptions.  What justifies this
"conservation of messages" assertion?

>          Added a multicast pattern (#8) that (hopefully) is what Amy
>          described

Well, I'm at a bit of a loss, here.  If #8 is what I described, then
what's #7?  If #7 isn't what I intended to describe, then what use is
it?

The intention of #7 was to describe a pattern in which:

1) the service sends a single message
2) the (zero or more) recipients of that message may choose to respond

with the fault behavior: any message may trigger a fault, which is to be
returned to the sender of the message triggering the fault.

Given that IOPs model only direction, sequence, and cardinality (if we
simplify by leaving out faults, for the moment), that looks like:

OUT, IN*

The agreement in Scottsdale was that faults go in.  That makes the
pattern more complex, but the core remains the content model shown
above.

Given the interpretation, in the MEPs created for IOP #7, that there is
*one* recipient of a message, who is not one of the recipients, there is
no use at all in those MEPs that I can imagine.

Amy!
-- 
Amelia A. Lewis
Architect, TIBCO/Extensibility, Inc.
alewis@tibco.com
Received on Wednesday, 9 April 2003 10:34:34 GMT

This archive was generated by hypermail 2.2.0 + w3c-0.30 : Friday, 25 March 2005 11:17:39 GMT