[omniORB] Application Design Question

[jorgefm@cirsa.com]

Hi all,

I'm starting a new project and i want to communicate all the modules using
CORBA omniORB.
All the processes will be executed in the same machine and CORBA is used
like an IPC
system to communicate different process, but in a future, it could be
possible to have
client's modules in a different machine and it will be possible if i used
CORBA from the begining.

I have some critical modules like the ErrorManager that I have to ensure
that always is alive
and the client's modules who wants to report a fatal error always will be
able to do it. In an
embedded system you always have to process the errors,and with fatal errors
you have to
be able to shutdown the system in a controlled way, etc even thought you
have some hardware fail.

Then, is there a way to make the omniORB using local IPC system calls (UNIX
sockets, named pipes, etc)
instead of TCP socket ? any command in the argv, etc ? I looked into the
code and the only thing i found
was the "-ORBendpoint giop:unix:<filename>" and "-ORBendpoint
giop:fd:<no.>". Anybody knows
what's the meaning of these options ?

I'm wondering if that way is more efficient that always use TCP sockets
because you can bypass the TCP
stack, the network driver, etc, but above all I'm looking for a robust way
to communicate errors.

Any comment, idea, design example, etc will welcome,

Thanks,
Jorge Fernandez