Re: requesting official list of modules and versions for GNOME 2.14

[Federico Mena Quintero <federico ximian com>]

On Thu, 2006-02-09 at 09:10 +0100, Vincent Untz wrote:

>   + glib + pango: the only objection was Federico's gripe about the
>     floating reference in glib 2.9. Federico, do you have an update
>     on this? Most people seemed to be happy to go with the new versions
>     (new stuff is gslice, pango/cairo and unicode 4.1).

Floating references went in, and I still think they are a terrible idea
for the reasons I wrote about in detail:

http://mail.gnome.org/archives/gtk-devel-list/2006-January/msg00012.html
http://mail.gnome.org/archives/gtk-devel-list/2006-January/msg00051.html

You have to understand floating references in the context of their
original purpose.  Quote:

	The complicated rules about GtkWidgets and their `floating'
	flag are there to avoid breaking *all* existing code.

[From http://mail.gnome.org/archives/gtk-list/1997-November/msg00245.html ]

Floating references were added to GtkObject to avoid modifying *all* the
apps written for GTK+ when we introduced reference counting.  Today,
putting floating references at the glib level is just a fetish for
gratuitous complexity.

Right now, my objection to floating references in stock glib is not that
of a technical problem --- I think even the ABI issues with the original
patches got resolved.  [Can we get *real* confirmation on that, by
someone who runs 2.12 language bindings with glib HEAD?  Otherwise we
are fucking ourselves in the ass very hard.]

My objection is that floating references introduce a consistency problem
for new APIs, a documentation problem, and it is just more pain for the
average programmer who wants to learn our platform at the C/C++ level.

Floating references do not help our users.

Floating references do not help programmers, either; they just confuse
them.

  Federico