On Fri, 29 Oct 2004, Carsten Ziegeler wrote:

Giacomo Pati wrote:
So, the question remains. Either we

   integrate & refactor the excalibur event package

or

   make the Quartz/Cron block core as it does all we need

Yepp, I think we should use a compromise here: let's take the
fastes road and add quartz to the core. The implementation is
hidden behind interfaces, so if in the future we come to the
conclusion that this dependency is not good for us, we can
just write our own implementation while preserving the interfaces.

And eliminating ALL spawing of new Threads in the code will eventually allow us to be more J2EE spec conformant (I do know of prominent J2EE app servers Cocoon isn't able to run in just because of spawning own sub threads).


I think this is on of our core problems: we not only rely on external
implementations but also on external interfaces. Changing implementations
is easy, changing interfaces not.

Yes, the JobScheduler is IIRC totally hidden behind our own interfaces (I'm the one who initially wrote it).


So, let's use quartz, but add a good interface above it (perhaps the
one from the cron block or a simpler one - i don't know); then refactor
our code to use this interface.

I'd propose to move the hole block as it is alreaday shielded behind our own interfaces. IMHO it doesn't make much sense to have two scheduling mechanisms availablebased on the same external dependency.


--
Giacomo Pati
Otego AG, Switzerland - http://www.otego.com
Orixo, the XML business alliance - http://www.orixo.com

Reply via email to