on 1-08-23 06.44, Andreas Schaefer at [EMAIL PROTECTED] wrote:
> When a client of an EJB can use RMI to access an EJB is it
> then also true that the EJB can use an RMI callback object
> (sent by the client) to send information back to the client ?
Yes - is my guess, - but - I do not know ... and will leave it to the people
in the know ...
> Let's say the client want to register an Event Listener on the
> EJB where then the events are sent back through the RMI call-
> back to the client.
A EJB follows a RRP (request / reponse pattern) as do Servlets !. To have a
keep-alive connection, from either, is a really exspensive way of trying to
achive a statefull / stateless connection with client (a push to client) !.
The application server has a managed instance pool of the above structures,
since the creation / destruction and each live instance of them is very
expensive, therefore it is bad to set a EJB or Servlet to be tied to one
client only, for any lenght of time but the absolute minimum, unless Your
pool is limitless ! or You *ever* only going to have just so many - as Your
pool is sized for ? ... Now lets assume You only have one client instance
that needs the keep-alive session, even then I would say, it is bad design
to use a EJB !. I mean I would not write a EJB for one client instance -
overkill ! ... code / deploy resource wise ? ...
The power of EJB is the instance pool, CMP, TX and scalability capabilities
... with the limits of RRP ...
Maybe You can send back a JMS handle from EJB to client ? - start and stop a
que throught the EJB ?
/peter_f
_______________________________________________
Jboss-development mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/jboss-development
