> The j2ee specs effectively require you to use thread locals in
> several places, and indeed I haven't found any way to implement JTA
> transactions without a thread local. It doesn't mean it's a good
> idea: for JTA the requirements basically come from C programming
> circa 1990 (maybe earlier). Jim has argued against thread locals
> repeatedly and to the extent I understand what he is talking about I
> completely agree with him.
>
> Here, I don't see why you would even think of thread locals. You
> have some method that may need a resource, why not just give it the
> resource to start with rather than forcing it to look it up? Even
> EJB 3 is moving towards some dependency injection/IOC concepts.

Yes, this is also possible. I personally am ok with TLS, but passing
an additional parameter to do callbacks is also possible. However this
would require us to talk about the callback mechanism too, which I
wanted to avoid. Seems that this is not possible, so a proposal for a
callback mechanism is needed, and it should be Vulcan-compatible to
get Jim into discussion.

Roman