On Tue, 2005-06-07 at 09:48 +0200, Stefano Bagnara wrote:
> I would also add to the list JMS (ActiveMQ) spoolmanager implementation:
> this would be a cool project and a big testbed for major changes in James 3.

I too would like to see this.  I think moving all the spooling over to
JMS would be a huge plus.  It would solve the clustering problem as the
issue would move over to the JMS provider (and ActiveMQ has a number of
solutions for clustering.)  It would add flexibility to James in that
users could use whatever JMS implementation they want.

The biggest issue I see with moving to a pure JMS based spooling system
is coping with failed message sends.  AFAIK, you can't schedule a JMS
message to be sent after a specified delay.  Perhaps Quartz could handle
this?

Noel keeps bring up the concern that JMS would be too slow but I'm not
convinced this is a issue.  My own tests have shown that ActiveMQ is
pretty fast and has a number of options for tuning performance.  I can't
see how JAMES' spool using a JDBC back-end would be faster than
ActiveMQ.  Of course, I could be way off base.  This is just my opinion.

-Mike



---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to