Re: Object Reference counting idea.

[Maciej Stachowiak <mjs eazel com>]

Michael Meeks <michael helixcode com> writes:

> Hi all,
> 
>         Here is my proposal ( based on a discussion with George Lebl in
> Toronto about what happens in the panel ); which is simple, almost
> entirely foolproof, trivial to implement and yet gives us a fair whack of 
> protection. 
>   
>         We partition refs and unrefs into 2 sorts; local and
> remote.
> 
>         * Remote is defined as coming through impl_Bonobo_Unknown_ref.
>         * Local is defined as coming immediately though bonobo_object_ref.
> 
>         NB. 'Remote' calls thus can happen in-process for shared
> libraries, or even by doing Bonobo_Unknown_ref/unref in the server
> process.
> 

This sounds much more feasible to implement, but I have a few questions.

1) Is the initial ref local or remote?

2) Does `query_interface' grab a local or remote ref?

3) What about other methods that implicitly ref the return value?

4) What about calling bonobo_object_ref on a BonoboObjectClient?


The other issue is that this approach makes the two kinds of refcounts
explicit and visible to the user. The confusion among gtk_object_ref,
bonobo_object_ref, Bonobo_Unknown_ref and CORBA_Object_duplicate is
already pretty bad. Right now, at least, bonobo_object_ref and
Bonobo_Unknown_ref do the same thing. If they were different, there
would be four kinds of refcounting for Bonobo objects, one of which
you should never use (gtk_object_ref). Is four (or three if you want
to call it that) kinds of refcounts for one object a good, clear
design? COM has only one.

 
>         Of course; if we try to do anything more for Gnome 2.0, I'd like  
> to read up on distributed garbage collection, everyone keeps claiming this
> is an unsolved problem and clearly there would be a load of issues. If    
> anyone has some nice links on the subject I'd much appreciate them.

I will try to look up some useful research papers for you.

 - Maciej