Trying to undestand container interoperability

Question

Hello !
After a while of reading the different j2ee specs (EJB, JTA, JTS...), I
thought I understood that it is possible for a bean in a container A to
invoke a bean in a container B. The dialog between containers is supposed to
be in RMI-IIOP.
First question : how does it look like from the point of view of the calling
bean ? does it perform a classic lookup like a normal EJB client ? In this
case, the lookup must be performed on the JNDI server of container B no ?
thus it is necessary to specify parameters when instanciating an
InitialContext, no ?
Second question : If the way the calling bean calls the bean in container B
looks like what I wondered in my first question, how can transaction and
security context be propagated to the container B ? Actually, if the bean A
retrieves a stub from the JNDI of B and performs invocation of B, how can
the B stub know that A is currently in a transaction in order to tell it to
B ?
Actually, while writing my questions, it looks like the B stub does not
necessarily talks to container B by using RMI-IIOP. So, what bean A should
do to make container A use RMI-IIOP to talk to B ?
Forgive me if my questions are a bit confused, but I desperately want to
know if and how it is possible to make two beans interoperate between
different containers. And as per the EJB 2 spec, RMI-IIOP seems to be the
solution, I'd like to understand how a bean can make use of this.
Thanks for your patience
Sylvain

Answer 1

Thanks for your reply. I still have a couple of questions...
"Andy Piper" <[email protected]> wrote in message
news:[email protected]..
"Sylvain Laurent" <[email protected]> writes:
Correct, but the lookup actually happens via CosNaming. So the scenariois:
>
Client -> JNDI -> CosNaming -----------> IIOP ----------> CosNaming ->JNDI -> Server
Your figure makes me think that the client instanciates an InitialContext
with the default parameters, no ?
And it has ejb-ref specially declared in its ejb-jar.xml ? Could you provide
me with an example ?
Transactional characteristics are specified in the exported IOR's and
transactions are propagated via OTS.You mean that container B exports an IOR to container A, which in turns will
know how to handle transactions when bean A invoke bean B, correct ?
Because in the case where A makes a direct look-up of bean B with the JNDI
of B, container A has no opportunity to know what's going on, correct ?
>
Actually, while writing my questions, it looks like the B stub does not
necessarily talks to container B by using RMI-IIOP. So, what bean Ashould
>
Why not?Because, for instance WLS uses its own t3 protocol. Thus I can say that if B
is WLS, B stub talks with container B using T3, unless I made a special
thing to force it to use RMI-IIOP.
Forgive me if my questions are a bit confused, but I desperately want to
know if and how it is possible to make two beans interoperate between
different containers. And as per the EJB 2 spec, RMI-IIOP seems to bethe
>
Of course. This has been demonstrated at the last JavaOne.Unfortunately I was not at the last JavaOne. What was the config for the
demo ? What containers ?
Actually I find it quite frustrating not to be able to find any page showing
a kind of step by step procedure to make two containers work together. Do
you know of any such a page ?
Thanks for your support.
Sylvain