Re: #101293, again

[Evan Martin <martine danga com>]

On Wed, Dec 10, 2003 at 10:30:40PM -0600, Gregory Merchan wrote:
> On Wed, Dec 10, 2003 at 10:04:12PM -0600, Gregory Merchan wrote:
> > On Wed, Dec 10, 2003 at 09:35:10PM -0600, Federico Mena Quintero wrote:
> > <snip>
> > > 1. Remove dialog_has_cancel() from gtkdialog.c
> > > 
> > > 2. Bind Escape to the "close" signal.
> > > 
> > > 3. Make the default handler for "close" just close the window with the
> > > existing code that synthesizes a GDK_DELETE event.  "close" is already a
> > > G_SIGNAL_RUN_LAST signal, so recalcitrant windows which really don't
> > > want to be closed with Escape can easily connect to the signal and stop
> > > the emission.
> > > 
> > > May I do all of these?
> > 
> > No.
> > 
[...]
> In short, window management is a mess. There should be an explicit
> moratorium on all feature changes until the "not built to spec" bugs
> are worked out, and apps should follow the specs. Twice broken isn't
> fixed. But, eh, good luck.

I'm in Federico's camp too, but I'm also sure there's good reasoning
behind your "no".  But the point you make here seems to be orthogonal:
there are bugs in window management, sure, but the Esc=Cancel binding is
at the GTK layer: the same place Federico's change would happen.

Could you discuss, from a "good UI" perspective (for everyone's
benefit, not just mine!) what *ought* to be happening here, in the form
of "When I hit escape, the [window manager / dialog code] ought to
[...].  The place where instant-apply windows are broken is [...]."?
(Or direct me to the relevant docs?)

-- 
Evan Martin
martine danga com
http://neugierig.org