W3C home > Mailing lists > Public > public-ws-async-tf@w3.org > March 2005

Does WSA in fact support arbitary interactions?

From: David Hull <dmh@tibco.com>
Date: Wed, 09 Mar 2005 16:33:42 -0500
To: public-ws-async-tf@w3.org
Message-id: <422F6BB6.9060205@tibco.com>

I've raised a question [1] on the WS-A list as to whether MAPs, as 
defined, currently support arbitrary interactions, as opposed to only 
supporting request/reply and patterns that can be bent into the 
approximate shape of request/reply.  For concreteness, I gave an example 
in which a node receives a message and may forward it to one of two 
destinations for normal processing, or may fault.

As far as I can tell, there is no way to specify these forwarding 
endpoints as message addressing properties.   The [fault endpoint] would 
be the fault destination, but what about the other two?  They can't both 
be the [reply endpoint], and in fact neither one should be since the 
intent is to forward for further processing, not to reply.  My take on 
the WS-A charter is that it requires support of arbitrary interactions, 
including but not limited to the one I described, on an essentially 
equal footing with request/reply.

Following this through, it would appear that the core spec has no 
business saying anything at all about request/reply, beyond possibly 
pre-defining some endpoint properties and relationship types in 
recognition that request/reply is a prevalent pattern.  Everything else 
should be, and to a great extent already is, discussed in the binding 
documents.  In particular, all discussion of default fault destinations 
and the anonymous endpoints should move out of the core.  I've also sent 
mail [2] to the list to this effect.

My question to the group is, how important are these issues?  As far as 
I can tell, the first issue, of whether MAPs really do support arbitrary 
interactions as advertised, is crucial.  Fortunately, it doesn't appear 
that hard to fix [3], if needed, but even if we don't adopt such a fix, 
we need a convincing story on how to handle interactions like the one I 
described, which may arise as part of larger choreographies.

Received on Wednesday, 9 March 2005 21:43:06 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 19:48:42 UTC