> Well, now I'm agreeing with you more ;-)

wow!! :)

> I'm reluctantly coming to the conclusion that to make the driver
> jca-compliant we have to fix this so autocommit = true if no other
> transaction control has happened.

that would be nice

> This is probably a good idea too. I think perhaps the best solution is
> another DataSource class that you can set the properties on (the

properties

> now in ManagedConnectionFactory) and that will create and set up the
> ManagedConnectionFactory, and make the ManagedConnectionFactory use it for
> its DataSource.

Yes, good idea

> What I don't want to see happen is the jca architecture compromised to
> support the Driver class.

that must absolutely not happen, and that's not what I was pointing out
about the actual implementation state.

> Adding conveniences such as the javabean
> DataSource that do not obstruct the jca framework is not a problem for me.
> (Not sure when I might do it, but even I might write it if I get a little
> free time ;-)

Well, I would like to say 'I'll do it', but actually my understanding of the
whole is still too poor, but maybe some time ...

Thank you
Massimo