From kde-devel  Wed Jun 16 16:48:25 1999
From: Lotzi Boloni <boloni () cs ! purdue ! edu>
Date: Wed, 16 Jun 1999 16:48:25 +0000
To: kde-devel
Subject: Re: Different corba implementations
X-MARC-Message: https://marc.info/?l=kde-devel&m=92955075825463

> >   Question: how difficult would be is to keep the system ORB independent?
> > The commercial ORBs like Visigenic and Orbix might be already in use on 
> > certain systems, and they might be better under some conditions. (Both of
> > them are considering Linux ports - it seems that Visigenic actually has
> > a beta). 
> > 
> >   But maybe linking against these would totally break the GPL so the whole
> > question is meaningless. 
> > 
> The question is meaningless, but not for this reason :
> 
> ORBs implementations are SOURCE incompatible.
> The CORBA so-caled standard is only a _specification_ standard, not an
> implementation one, so each ORB is free to implement things
> as it wants.
> This means you can't develop for all ORBs at the same time, you
> have to choose one - but it can interoperate with the others since
> the protocol (IIOP) is part of the specification.

  Hmmm... 

  I think that they ARE source incompatible, but not that MUCH. Large
books like "Advanced CORBA programming with C++" are presenting thousands
of lines of CORBA code without mentioning which implementation you should
use. And it works with MICO. 

  I will find out soon how big is this problem, because I will port my
code from MICO to Orbix and probably to Visigenic too in the following
days (an AT&T research project). We had some phone conferences with the
companies and I harassed them into porting to Linux :-). They are rather
anxious not to loose a consumer like this. I directly asked
them about source compatibility and both companies answer was that there
will be problems at "details" - I guess memory management is the first
one. 

   And BTW, about 10% of this AT&T lab is running Linux and KDE.

           Lotzi B.