What is wrong with the lifetime parameters?
Can't you just set these to a high number?
Lifetime parameters are fine. But whatever values I would choose would simply be a guess. If a user leaves my app up over lunch, over the weekend or for days on end, it would be nice if his SFSB were around when he returned. It's important that I keep the long-idle users as happy as the ones who use the app for only a few minutes. My fear is that by choosing, say, a large max-bean-life that I may negatively impact the timely reclamation of objects in JBoss.
That was my point about RMI. What I like about it is that it does reference counting. A remote object knows when it's no longer referenced and won't be gc'ed before that time. I thought if it were somehow possible to incorporate that aspect of RMI with the part of EJBs I like most, passivation, that would be the best solution for my application.
Thanks for your input.
There is no direct link between the SSB instance and
the client. It is the remote interface that is an RMI
Why don't you just specify a reasonable passivation
period (the bean will go onto disk) then leave it there
for 24 hours, 2 weeks or whatever?
> There is no direct link between the SSB instance and
> the client.
This is the point I was missing. If there's no link, then obviously my hopes for something akin to distributed garbage collection will never work. Thanks for the info. I guess I'll take your advice and set a reasonable timeout for the bean life.