Re: Steps to get to GTK+ 3.0

[Cody Russell <bratsche gnome org>]

On Wed, 2008-06-04 at 19:59 +0200, Jean-Yves Lefort wrote:
> > I fail to see how any of this discussion should change your
> perception
> > of GObject.  This is a discussion about GTK's usage of GObject, but
> your
> > own objects can continue to use it in whatever way makes the most
> sense
> > to you.
> 
> Tim Janik argued that the bad performance of the property system is
> not an issue in GTK+. I pointed out that in my understanding, the
> GObject framework is supposed to be a client-agnostic object
> system. As such, its performance issues should matter to the GLib
> maintainers, regardless of how much they affect GTK+.

Right, but that's a different discussion.  Nobody is talking about
slowing anything down GObject, so I didn't really understand what your
objection was to in particular.

The property system is slower than directly accessing a field in a
struct.  When this was pointed out, my understanding of Tim's response
was that we will only use the slower method in situations that are not
time-critical and will yield benefits in other ways such as future
maintainability of binary compatibility.

If there are ways to speed up the property system that would be
fantastic, but that should be discussed separately I think.

/ Cody