Re: GnomeClient replacement?

[Bill Haneman <Bill Haneman Sun COM>]

On Wed, 2006-07-19 at 17:01, Shaun McCance wrote:

> Hey Bill,
> 
> As usual, I'm afraid most of us don't understand all the layers
> as well as we ought to.  Could you clarify exactly which pieces
> of the accessibility stack wouldn't get activated?  There are
> a lot of GTK-only applications, probably a couple inside Gnome.
> What are these applications missing out on?

For now, as "pure gtk+" apps are relying on gnome-session to set the
GTK_MODULES environment variable to load the modules "gail" and
"atk-bridge".  This is not really optimal but it works well enough for
gtk+-only apps (some apps that use gtk+, such as mozilla/firefox, have
to take measures to un-set this variable and do their own accessibility
work based on the gconf key).

What we really want is for all desktop apps to have a way of detecting
the need for assistive technology support without having the environment
tell them what modules they must load in order to get it - in this sense
the GTK_MODULES solution makes the apps accessible "as a side effect",
rather than providing clear semantics as does gconf.  A boolean XSETTING
would probably be nicer since it would allow apps to make their own
decisions on what modules should be loaded in order to enable assistive
technologies.  This would mean that gtk+ would incur a soft dependency
on libgail and at-spi which it currently does not have, but at this
stage in the history of ATK this would probably be acceptable.  It has
been proposed that gail be rolled into gtk+ at some point, which would
be very helpful (as a separate module, it can be very awkward to poke
into gtk+ widgets to the extent necessary in order to implement ATK on
their behalf).

The $GTK_MODULES approach also fails for embedded apps and things that
use gtkplug/gtksocket, since for those apps an additional module
(libgail-gnome) must be loaded - unless we don't mind pure gtk+ apps
loading that module needlessly.

regards

Bill

> --
> Shaun
> 
>