Re: oaf / bonobo patch ...

[Maciej Stachowiak <mjs noisehavoc org>]

On 23Oct2001 09:27PM (-0400), Michael Meeks wrote:
> 
> 	So it transpires the fix or workaround for the shlib activation
> factory issue is to pretty much ignore the return value from the oaf
> registration.
> 
> 	It seems to me there needs to be a more robust definition of the
> difference between shlib and out of proc components wrt. their
> registration and activation. 

As I said on IRC, shlib servers should not register at all in any way.

Currently, shlib servers have "implicit factory" semantics - if you
activate one server multiple times, you will get back different
objects.

So there is no point to registering.

If you really want a singleton object, you can make the shlib activate
function return the same object multiple times.

This is not necessarily the cleanest design (we could try to do
something better for bonobo-activation I suppose) but that's how it's
meant to work.

 - Maciej