W3C home > Mailing lists > Public > public-soap-jms@w3.org > August 2008

RE: RE : WS-Addressing

From: TALBOT Jacques (TJA) <Jacques.TALBOT@teamlog.com>
Date: Tue, 19 Aug 2008 11:58:29 +0200
To: "public-soap-jms@w3.org" <public-soap-jms@w3.org>
Message-ID: <A863DA9C49F2224A85129B1825E7B24B57C63A806A@TLGCEX01.teamlog.intra>
Disclaimer: I do not want to "bother" specifically this WG
Most of the dialog difficulties seem to derive from the fact that I am trying to talk to "WS-* as a whole", or at least to W3C as a whole; unfortunately, the relevant specs are spread on IETF, OASIS and W3C, and JAX-WS in java.net.
But behind the curtains, it is quite oftem the same people from the same companies and fortunately WSA and SOAPonJMS are all W3C.
As a  user, I seek guidance on how to assemble these pieces in a vendor-independent way, programming in JAX-WS.
ws-i.org has apparently failed to accomplish anything beyond the most obvious MEPs.
So, where whould I ask? I try here, reinforcing hopefully Nathan's questions:

SOAP on JMS draft:
    soapjms:replyToName
   the endpoint is according to the IETF JMS URI scheme

WS-Addressing standard:
  wsa:ReplyTo
  wsa:FaultTo
  + some specific semantics for anonymous

The WSA endpoints are under the JAX-WS (e.g.) programmer's control.
The soapjms port is under the SOAPonJMS binding implementor's control.
Where should it be explained how these abstractions are related?
Is it so obvious that soapjms:replyToName and wsa:ReplyTo are the same URI that it does not have to be mentionned anywhere?
What about FaultTo?
What about anonymous?

In a similar vein:
wsa:MessageID and wsa:RelatesTo are under the JAX-WS (e.g.) programmer's control.
JMSCorrelationID and JMSMessageID are under the SOAPonJMS binding implementor's control.

I understand from Eric's answer that these are unrelated. In a way, the binding code can "relate" them, but this is irrelevant for the programmers and the run times (APIs and protocols) on both sides of the connection.

>From your experts' standpoint, all of this is of course trivialities and should be left as an exercise to the reader :-)
>From my user's standpoint, it is better written down than assumed obvious.

A more political view
Pleeeease, be pedagogical and get out of the ivory tower or we will all go the EJB2 way; WS-* is badly suffering from its complexity, it becomes a hard sell for many projects. RESTful is so much more fun these days!

So definitely, an end to end description of how we can program in JAX-WS on SOAP on JMS an asynchronous on-the-wire Request Reply MEP would definitely help. My experience is that if you really TRY to program this and interoperate it between e.g. IBM WMQ and Tibco EMS, you will discover holes in the chain of specifications. (It is quite the case with the supposedly time weathered http binding)

Good luck to all
--
Jacques Talbot - Teamlog 10 rue Lavoisier - 38330 Montbonnot
Tél: 04 76 61 37 12  Mél: jacques.talbot@teamlog.com
Tél. mobile 06 07 83 42 00


________________________________
De : Eric Johnson [mailto:eric@tibco.com]
Envoyé : mardi 19 août 2008 07:36
À : Nathan Sowatskey
Cc : Phil Adams; TALBOT Jacques (TJA); public-soap-jms@w3.org
Objet : Re: RE : WS-Addressing

Hi Nathan,

Nathan Sowatskey wrote:

Hi all

Apologies for the abrupt silence. I am travelling at the moment.

My main concern would be to understand how the WS-Addressing headers map to
JMS headers.


For the most part, they simply don't.  That is, the JMS "correlation id" is irrelevant to the SOAP binding for WS-Addressing.  The SOAP binding for WS-Addressing puts the relevant information in the SOAP headers.

Further, we are not attempting to describe a non-SOAP WS-Addressing solution over JMS.

I suppose there is a small glimmer of ambiguity about the meaning of the "anonymous" back-channel, however, we could simply recommend in our FAQ that people use the JMSReplyTo header for that purpose.  Off the top of my head, I cannot think of any reasonable alternatives here, so a lack of a specific specification doesn't seem like a glaring oversight.

One question to consider - if there is a non-anonymous EPR (perhaps for faults), have we sufficiently defined the URI schema that it is possible to send the appropriate URI in the EPR for the fault in the request message?  My take is that we have covered that case with the JMS URI scheme (done separately).  However, it is also worth pointing out that for the JMS case, it is highly unlikely that you *need* to specify an alternate EPR for either ReplyTo or FaultTo, because chances are they're the same, and the requester can already pick an alternate JMS destination - meaning that you can use "anonymous" as the EPR indicator.

In other words, the only place you might get into any trickiness with WS-Addressing is when you're crossing transport notions - HTTP request with JMS response.  As I said, I think the JMS URI should handle this scenario.

-Eric.
Received on Tuesday, 19 August 2008 09:58:55 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Saturday, 18 December 2010 18:16:17 GMT