Re: Enter the build sheriff: Jacob.

[Seth Nickell <snickell stanford edu>]

> You are also too close to the project to see the problems. You must feed
> Gnome 2.0 final (ish) to people who are going straight from 1.4. You'll
> see deep annoyance at the dialogs without any understanding of why it
> irritates them. Take 30 people into a lab and try it some day.

I don't claim to be a good UI designer, but a good UI designer is one
who has learned to accurately predict user reactions before testing, and
hence can fix most of the problems before the test (of course, you
should always be testing to *confirm* that you are doing things right,
the point is that the tests will usually turn up refinements instead of
major problems). I just think its worth mentioning that being close to a
project does not preclude one being able to see *most*, if not all, the
problems. There will of course always be suprises in user testing.

Its like performance tuning. The bottom line is "does this run fast
enough" (or in little enough memory, or with so much latency or
whatever, choosing metrics can be tricky, of course, and the same thing
is true of user testing). And you measure that by a set of tests,
whether "real life" tests or some sort of benchmark, probably both. But
there is still place for a good designer. A good designer means that
there will be fewer serious kinks that have to be worked out when it
comes time to start optimizing. They have acquired a sense of what
architectures and patterns result in good performance. er, this is a
real rabbit trail though ;-)

I haven't done any substantial testing of the dialogue button order on
GNOME 1.4 users. I have taken careful notes on 6 GNOME 1.4 users as they
installed and first worked with GNOME 2.0 on their boxes. Nobody
commented on the dialogue order, and I did not notice any mistakes that
might be attributed to it. But this is terribly weak evidence,
particularly because GNOME 2.0 had, and still has, *much* larger
problems, so I suspect even a rather large problem caused by the button
ordering would have been obscured.

The closest testing I have is: 
http://lists.gnome.org/archives/gnome-list/2002-February/msg00328.html

Which is on Windows users (primarily), but it was not a formal test
intended to explore dialogue order. In my defense, I'd like to point out
that user testing takes a long time, I can only convince random people
in my dorm to "waste" so much time being tested by me, and I do a *lot*
of other things besides user testing.
 
> Most of what Seth has done seems to be in the right direction. I'd dearly
> like to see a post-seth sawfish capplet for example. 

http://beauty.stanford.edu/window-properties.png

I didn't have time to finish implementation of this for GNOME2. Maybe
for 2.2. It will have pluggable backends so different window managers
can support the settings exposed by the capplet (theme, font, and focus
method).

Ideally, I don't think the window manager should be an exposed setting,
but I suspect that many (though not most) existing GNOME users will want
the setting, even if they never change it, a consquence of window
managers being a hot topic in the past. I considered labeling it "Draw
window borders with [Sawfish     ]" but I decided if the setting is
going to be there primarily for existing users looking to change their
window manager, we might as well label it in the way they'll be most
likely to find it.

-Seth