Re: Installed IDL files vs CORBA stub/skel Libraries.

[Miguel de Icaza <miguel helixcode com>]

Hello guys,

   So we are running into a problem that we need to standarize soon.

   Typically we ship and install IDL files on a system, because this
is the published interface to a given component.   

   Sometimes we have libraries that need to talk to a CORBA server
implementing one interface, so we put the iface-stubs.c and
iface-common.c into a library so the library can talk to a server.

   Sometimes, the CORBA interface uses callbacks or notifications, so
we need to also ship the skeletons and provide an implementation for
those skeletons.  

   Sometimes, we might ship the CORBA stubs, to encourage code reuse,
and avoiding each application having its own copy of the stubs.

   Sometimes an IDL file includes another IDL file.  When the included
IDL file is supposed to install also a convenience library we have
hacks like '#pragma inhibit push' inside an ifdef and a '#pragma
inhibit pop' to deal with this problem.  The naming scheme for the
defines is not clearly defined either.  I have seen a couple of
variations over time. 

   We need to standarize this issue:

	1. Standarizing the define name for forcing compilation of
           interfaces.  

	2. Do we want to trigger compilation in a per-idl file or in a
           per-interface basis.  I think we want the later.

Best wishes,
Miguel.