David Ingham wrote:
Hi Rafael,
I think that the spec extract that you quote is addressing a different issue from how I understood Rajith's original ask.
It's actually from the javadoc for MessageConsumer.setMessageListener(),
not the spec. I wouldn't be surprised if the spec had more explicit
language on this subject.
I believe that the extract you quote below is talking about what happens when I consume messages via different routes from the same MessageConsumer. That is, if I set MessageListener, ML1, on MessageConsumer, MC, and I then I also receive messages synchronously by calling MC.receive() or I set another MessageListener with MC.setMessageListener(ML2) then I think the spec is saying that the behavior is undefined, i.e., how messages are distributed between ML1, ML2 and the synchronous receives is undefined.
I interpreted the original description as something different, namely I create
a Connection, create a Session, start the Connection and then create a
MessageConsumer with a MessageListener, i.e., the connection had no
MessageListeners at the time it was started and then they were latterly added.
My previous statement was that I believe that this sequence of events is
legitimate and these new MessageListeners should begin to receive messages. I
interpreted Rajith's original mail as saying that the implementation didn't do
this and I would classify that as a bug.
That may be a reasonable interpretation as well. I guess it depends on
what the term "undefined" applies to, i.e. the method semantics vs the
distribution behavior.
In any case it doesn't seem unreasonable to make the example work. I
would just caution against putting it forward as an example of a good
pattern to follow.
--Rafael
---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project: http://qpid.apache.org
Use/Interact: mailto:[email protected]
