Re: gnome-triggers/gnome-sound/sound events

[Havoc Pennington <hp redhat com>]

Bill Haneman <bill haneman sun com> writes:
> Elliot Lee wrote:
> > 
>  
> > When programs want to generate a .soundlist type of sound event (such as
> > "button clicked" or whatever) they ask gnome-session (via
> > gnome_client_app_event()) to make it happen.
> > 
> > So aside from s/gnome_triggers_do/gnome_client_app_event/, there's no
> > change from the app's viewpoint (and not many apps use this API anyways).
> > The main gain is from removing the dependency of tons of apps on libesd
> > and libaudiofile, to decrease startup time, memory footprint, and
> > dependencies in general.
> > 
> > -- Elliot
> 
> This has the added benefit of making it easy to support alternate,
> non-audio 'notifications' if the audio file is actually conveying
> (non-redundant) information.  This is good for accessibility for
> hearing impaired (and the audio-driver-impaired ;-)
> 

If we're going to start packing things in, then the part in
gnome-session needs to be kept bounded and minimal; maybe
gnome-session would just pass notification along to yet another
process. i.e. gnome-session could be just a switchboard, it doesn't
run the actual code that plays the sound or whatever.

Putting code in gnome-session is just not good when it can be avoided,
especially if said code is complicated. gnome-session crashing is
very, very, very bad and will drive people away from GNOME in a
hurry. (And a nightmare to debug, the crashes are a headache to
reproduce usually.)

Havoc