Rickard,
>If you do not talk through stubs, how do you intend to comply with
>section EJB1.1 B.9? (unless you wanna be clever, in which case weak
>references will help, yes)
Ejipt only exposes stubs to clients (both local or remote) hence it complies
with the spec. However, since locally stubs are all shared (by the JDK
implementation), your server cannot get unreferenced notifications when your
*local* clients (i.e. peer beans) release the local, shared stubs.
>True, but it will be mandated (AFAIK) that IIOP is *possible* to use,
>and since there is no DGC in IIOP this becomes the "worst" case.
I agree, at this point it seems that IIOP will be possible to use but not
all the goodies will be available.
>In any
>homecooked protocol (including JRMP) you could include DGC as you
>please, but not in IIOP (again AFAIK, CORBA is a big dark hole for me).
DGC is an advanced concept available only when you have local GC. Mainstream
CORBA languages (C, C++) do not allow GC. Well, I think I am getting
outside of my jurisdiction here...
>All this would be pointless *if* stubs were allowed to contain custom
>code such as:
>* on method call to stub, delegate to server
>* if ObjectNotFound exception thrown, call finder to get fresh instance
>and retry
>but since (AFAIK) CORBA clients create their own stubs this aint
>possible.
You still need a runtime environment.
Imre Kifor
Valto Systems
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff EJB-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
