Re: On the BoF agenda, user interfaces, and developer resources

Joe Harrington (jh@tecate.gsfc.nasa.gov)
Thu, 19 Oct 1995 18:23:58 +0100

My concern is that by not starting with scope and basic user
requirements, we will decide it's ok to take some paths that it isn't
ok to take. Previously on this list people have proposed implementing
directly in current Unix standards rather than having a portability
layer. This makes a number of assumptions, some of which are not
immediately clear. Had we first discussed user requirements and (for
example) agreed that Unix is not the only OS anymore, those arguments
would not have held much water.

I have no illusions about defining how people should want to work. I
only want to come up with a list of requirements that people will and
already do have, so we don't accidentally ignore an important one and
doom the system from the start. Here's where I'm coming from. When I
was a student, I worked in the Systems Development group at MIT
Project Athena. Others were writing the X window system at the time.
X was the dream. X would solve all the problems. X would do so much
and so well that everyone would adopt it and it would become the
standard. It would be free and it would grow. It would support any
look and feel. Somehow, through all that, they forgot to take into
account what it would be like to use it, and to program in it. You
can do an awful lot in it, but most people I know who used to make
great graphics programs on their pc's at home (before the IBM PC stole
the name) have never learned how to do it in X. It's too complicated.
Most secretaries and a lot of scientists I know fear to learn *even
the user level* because it's too complicated. Now people build
toolkits and "added value" vendor interfaces on top of the mess to fix
the problems, but the performance hit is large and people still don't
really like them (remember when a Vax 750 with a VS100 display ran the
window system just great? not anymore). I'd like not to see this
happen again.

--jh--