Hi Osman,

Thanks for your input.  Sorry for the slow response.

Osman Mrzljak wrote:

Dear editors,

 

I just read “SOAP over Java Message Service 1.0”. I am very happy with this working group which will deliver standard for JMS transport for SOAP messages.

 

All specifications (I found) at this moment are vendor specific on some why. In many usages of JMS as transport I saw not correct usage of JMS properties in WSDL. I think this draft is making the same mistake.

 

From my point of view, a WSDL file:

-         describes a services interfaces and binding …

-         is developed and delivered by Servcie Provider.

-         doesn’t contain  any portion of data specific for one of clients (Service Consumer). So no client specific information in WSDL. This is only way to deliver one WSDL and publish it in Service Repository. If client information is embedded in WSDL, Service Provider has to deliver one WSDL per client (Service Consumer).

 

In your specification, there is possibility to specify following properties in WSDL:

-         jndiConnectionFactoryName

-         jndiInitialContextFactory

 

Both are client configuration.

 

I see following problem: Service Provider can’t develop one WSDL for all Service Consumers. Service Provider can’t develop one WSDL per client because he/she doesn’t know client specific configuration. Or is your idea that Service Provider develop WSDL without JMS binding specified and any Service Consumers (clients) add it’s JMS binding to it’s version of WSDL file.

You have pinpointed a key issue with any JMS binding - how do you specify connection information?  The working group has come to the conclusion that it may be useful in some scenarios to specify the JNDI information, although it is clearly not generally useful in all circumstances.

What we've tried to do then, is supply a way that someone using the SOAP/JMS binding in WSDL can use to specify JNDI information, but they do not have to.  That is why in section 2.2.1
(http://dev.w3.org/cvsweb/%7Echeckout%7E/2008/ws/soapjms/soapjms.html?content-type=text/html;%20charset=utf-8#binding-connection), we state that the JNDI information MAY be specified, but it need not be.

If you are hoping that we would define an generalizable mechanism for discovering JMS endpoints, that work is out of scope for the committee - we're defining the specification to work with existing approaches.

-Eric.

 

Thank you very much for this specification work.

 

Osman Mrzljak

www.osmanmrzljak.com