[omniORB] Question about socket connection management in ORB

[Craig Rodrigues]

    TAO VERSION: 1.1.14
    ACE VERSION: 5.1.14
    OMNIORB VERSION: 3.0.3
    OMNIORBpy VERSION: 1.3

    HOST MACHINE and OPERATING SYSTEM: Redhat Linux 6.2
    COMPILER NAME AND VERSION (AND PATCHLEVEL): gcc 2.95.2

    SYNOPSIS:
     General question about how the ORB manages socket connections.

    DESCRIPTION:

I am developing a simple testing application used to test the IDL interface
exposed by C++ TAO CORBA servers.  My test application is
written in Python, and uses the Python CORBA bindings in
omniORB: http://www.uk.research.att.com/omniORB/omniORBpy/.
The CORBA language mapping for Python is very simple (20 pages!), and I have
had great success in using Python as a language for developing
test scripts for C++ TAO CORBA servers.

In my test Python application, I get an object reference to the
C++ TAO CORBA server, invoke a bunch of methods on this object reference
(which are exposed in the IDL), wait for a few minute or so, then
terminate the test application.

If I increase the TAO ORB debugging with -ORBDebugLevel 1, I notice
that when I terminate my test application, the ORB on the C++ server seems
to close a socket.  This is true even if I terminate the test application
a minute or two after the last invocation of a CORBA method.

When a CORBA client and server connect via IIOP (TCP), is some kind
of connection caching/reuse going on?  Is this an artifact
of TAO or omniORB?  Can it be configured?  I can see some performance
advantages for doing this vs. making a new TCP connection every
time a new CORBA method is invoked on the same object reference.

Thanks for any clarifications.
-- 
Craig Rodrigues        
http://www.gis.net/~craigr    
rodrigc@mediaone.net