Very interesting. These would indeed require very limited changes to the
JCA spec (if any).
What do others think about David's proposal?
David,
Couple of questions/comments:
1. Can you change your example to call createEndpoint() just before the
connector calls the bean method (and release the instance after the
call)? This will fit much better with the existing rules on the MDB
lifecycle (and reduce the amount of changes)?
2. Can we skip the marker interface altogether? If there is no
implementing interface, it's a no-interface view. Plain and simple.
3. Use the "beanClass" instead of the "ejbClass" for the property name?
thanks,
-marina
David Blevins wrote:
> Here is a better description of the revamped proposal.
>
> https://github.com/dblevins/mdb-improvements
>
> Note, this has nothing to do with JMS or JMS 2.0. This Connector/MDB enhancement has merit in its own right.
>
> The proposed enhancement is at the bottom of the page:
>
> https://github.com/dblevins/mdb-improvements#message-driven-beans-tomorrow
>
> The example comes with a working Telnet Connector that can be run via Arquillian. Just hacked it up to demonstrate a non-asynchronous mindset and show the potential of what is already there and can be done today. Generally to get the creative juices flowing. Should run in any server.
>
> https://github.com/dblevins/mdb-improvements/blob/master/mdb-today/src/test/java/Runner.java
>
> I think we can get a lot of bang for such a small tweak. For all of us with existing MDB Containers, this should be very easy to slot in. Low impact, high value.
>
>
> -David
>
>