Re: [Usability] Re: Error dialogs (among others?) and focus stealing prevention

[Elijah Newren <newren gmail com>]

On Wed, 27 Oct 2004 12:03:25 -0400, Havoc Pennington <hp redhat com> wrote:
> On Wed, 2004-10-27 at 08:25 -0600, Elijah Newren wrote:
> > Interesting idea.  I'll try to do some tests locally to find out.
> > There also appears to be another half to the question, though, that I
> > think is important:  How many windows that are supposed to get a
> > timestamp of 0 (i.e. unexpected windows) will end up getting a real
> > timestamp if we try to rely on the current event rather than a hint
> > from the application?
> 
> My view is that failing to focus windows that need focus is a worse evil
> than focusing some "out of the blue" windows, if it comes down to a
> tradeoff there, I would say focus by default and require apps to set the
> timestamp to 0 to avoid it.

I've been busy, but I finally got around to do a little testing of
your idea.  Unfortunately, it clearly presents a tradeoff.  So, as you
say, I think that means we should avoid this gtk change and just have
apps set the timestamp to 0 to avoid getting focus when needed.

(More details for the curious: I only tested nautilus; it seemed to be
a good enough test case and one which I thought would span the
spectrum of things I needed to test.  And the change totally busts
it--many windows which should be focused aren't.  nautilus apparently
queues an event whenever a user (double) clicks on an icon to open a
new folder or application or whatever, and when it actually starts
processing that queue, the event that caused it is no longer the
current gtk event.  The same is also true for key navigation such as
alt+up.  This means that the current event timestamp is 0 and metacity
treats the newly opened window as though it requested to not be
focused.  In spatilus, this just isn't good--each new folder is
obscured/hidden as the user digs down a hierarchy of folders.  We
could force nautilus to change and record timestamps and always
manually set them, but that seems to be requiring work to get the
common case right merely for the purpose of trying to make the niche
case automatically right; as you said that seems to be backwards and
the common case should be automatically right while the niche case
should require more work.)


Cheers,
Elijah