Re: Committed patch to bonobo-activation

[Maciej Stachowiak <mjs noisehavoc org>]

On 18Oct2001 11:50AM (+0100), Bill Haneman wrote:
> Maciej Stachowiak wrote:
> ...
> 
>  Also, it's probably not a good idea to try to access the stuff in
> > Bonobo_ActivationContext.idl or Bonobo_ObjectDirectory.idl. The
> > bonobo-activation library handles important parts of the work,
> > including launching the server and loading shared library
> > servers. These files should not be considered a public CORBA interface
> > to the bonobo-activation server.
> > 
> > Yes, it's a bug that there is currently no public CORBA interface to
> > libbonobo-activation. I'm hoping that will be fixed for 2.2.
> 
> Ah, trouble is, we need it *now*... 

It would be nice if I had known about this need before the API freeze,
say around the time I posted a list of future plans which said a
public CORBA interface was a post-2.0 item.
 
> Any chance we can get a subset agreed to before FCS?  We need to provide
> something that uses it...
> Probably just activate_from_id would be enough.
> 
> Our interoperability between the Java Accessibility API and the Gnome
> Accessiblity API requires it, and we really don't want to ship a Java
> engine that depends on bonobo-activation binary internals...
> 

When there is a public CORBA interface, it will almost certainly be an
in-process component and will require a C function call to get
it. Otherwise, you have an infinite regress of bootstrapping
problems. If you know of a better way to get at a primordial CORBA
server that can be used for activating others, I'm all ears.

Another approach is to replicate the libbonobo-activation logic in
Java (as MonkeyBeans did), but this is risky and unsupported.

Anyway, if we can come up with a good design and it's ready before the
new "we really mean it this time" API freeze date, I'll consider it.

 - Maciej