On 12/01/2012 15:43, Nigel Deakin wrote:
> Dear expert group,
>
> This email is, as usual, addressed at the expert group and subscribers to the user alias. However I'd particularly like
> to address it to those who develop JMS implementations. (You know who you are :-)
>
> I refer to this JIRA issue:
> http://java.net/jira/browse/JMS_SPEC-45
>
> We've discussed this before, and I'm now revisiting this issue with the intention of drafting suitable changes to the
> spec and javadocs, and in proposing suitable clarifications to propose to the EJB 3.2 spec.
>
> Please read that issue for a reminder of the this issue.
>
> That issue explains that createSession() has two arguments when it really should have only one, or even none. However
> we've already discussed what to do about that and I'm not concerned about that right now.
>
> My area of particular concern today is the behaviour of connection.createSession in various Java EE scenarios. This is
> described in two sections of the EJB 3.1 specification: section 13.3.5 "use of JMS APIs in Transactions" and section
> 13.6.5 "Handling of Methods that run with an unspecified transaction context". As my analysis in that issue explains,
> these sections are not very clear and leave several questions unanswered.
>
> I think we should attempt to clarify this for JMS 2.0 and EJB 3.2, in the javadoc for createSession and also elsewhere.
>
> However there is a danger that if we attempt to "clarify" something which is currently underfined, ambiguous or unclear
> in the spec we may unwittingly cause JMS providers to change the way they behave in a way which will cause existing
> applications to behave differently.
>
> So I'd like to ask all the JMS vendors out there: how does *your* JMS provider interpret these two sections of the EJB
> spec? (Note that the Java EE spec states that they applies to the web container too).
>
> Here are some of the questions I think JMS vendors need to be able to answer, but which are not unambiguously answered
> in the EJB spec:
I've tided up these questions and added a new one relating to the use of "regular" JMS connections in Java EE, and added
them as a comment to the JIRS issue.
Here's a direct link
http://java.net/jira/browse/JMS_SPEC-45?focusedCommentId=328983&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#action_328983
As we answer these questions, or discover how existing application servers answer them, I'll update that comment.
Nigel