Re: Proposing GtkUnique 1.0 as a blessed external dependency

[Emmanuele Bassi <ebassi gmail com>]

Hi Paolo;

On Tue, 2006-11-14 at 13:43 +0100, Paolo Borelli wrote:

> > GtkUnique is now hosted on GNOME CVS under the gtkunique module[2] and
> > if needed I'll also request a Bugzilla product for it; I can commit it
> > to API+ABI freeze after the 1.0 release - which should happen as soon as
> > the end of the month.
> > 
> > Any comments are, as always, welcome.
> 
> How big is it in code?

It's a public object (GtkUniqueApp) plus two enums (GtkUniqueCommand and
GtkUniqueResponse).  The IPC code is in the various backends; which
backend is going to be linked in is decided at compile time.

>  I know everybody hates libegg, but if it is
> little I think it would be not too bad to cut&paste it for now, since I
> *really* want this kind of thing to hit glib/gtk proper.

I would really like for this to get into GTK+ too; there are two issues,
though, which should be resolved:

  1. time frame: I cannot guarantee that such feature is going
     into GTK+ 2.12 in time for GNOME to reach API freeze;
  2. logical place in GTK+: since there's no "GtkApp" class
     inside GTK+, having a stand-alone GtkUniqueApp class
     makes much less sense - also considering that even if we add
     a GtkApp class later we cannot change the inheritance
     of GtkUniqueApp as being a GtkApp, for API/ABI guarantees.

Hence, GtkUnique in GTK+ makes sense only in the long run, when GTK+
will have a generic application class.

>  This would
> avoid yet-another-mini-lib that distro have to ship and then keep around
> for a long time because any random third party project may have started
> to depend on it.

A way to avoid having direct dependency on GtkUnique would be to use CVS
inclusion inside a host module, without literally cutting and pasting
code; the changes in the build would be really small.

> Even if it stays an external lib, can you make sure to not use the gtk_*
> namespace? This way we can eventually put it into gtk at a later point
> without symbols conflicts.

This can be fixed before 1.0, but given the number and usage of symbols
GtkUnique exports, I would expect them to be the same in case of
inclusions inside GTK+.

ciao,
 Emmanuele.


-- 
Emmanuele Bassi,  E: ebassi gmail com
W: http://www.emmanuelebassi.net
B: http://log.emmanuelebassi.net