Unix, IBM, humankind, Smalltalk, animation, parallelism
ucbvax!works
Please, let's not crank up the old Unix religious debate on WorkS. See
back issues of nearly every other digest on the net for similar debates.
I think it's clear that the critical resource in today's computing world
is manpower, not machinery. The result? Systems which encourage and
support software movement among a variety of hosts are bound to be popular;
Unix and CP/m fill that bill reasonably well (no comparison betwixt the
two intended).
Since this list is supposed to be a collective soul-search for the ultimate
workstation (whatever that is), what might conceivably serve as a basis
for the next wave, after Unix? I propose Smalltalk (ST80 for short).
ST80 sits just about on the edge of what today's hardware (read: 68K
equivalents and follow-ons) can support "reasonably". I've seen ST80
in use, and know that I'd much prefer that world to anything else I've
seen (including Lisp machines, which have perhaps the most complete, if
also the most arcanely complex programming environment today). I think
the XEROX Systems Concepts Group (or whatever they call themselves now)
also knows it (why else would they have devoted so many years to ST?)
It's a shame that they're having such a hard time releasing ST to the
world, but I hear they're on the verge of such a release after a long
battle to get licensing going. (I hate to speculate like this, but
I can't say all that I hear.)
ST80 has exactly the properties proposed recently: an extensible system
with an "animated" feel. It doesn't use parallelism in any large
way, but I'm convinced that we don't understand how to use massive
parallelism in any but the crudest sense (e.g., weather prediction).
My ideal workstation, perhaps buildable by 2000, would place me in a
virtual world (3d and all) where the hallmark was malleability of the
objects and actors in the universe. This is an old idea, but makes
complete sense to me. And no one is doing anything seriously along
these lines. Boils down to the same old problem: no one knows how to
use massive amounts of computing power. The old paradigms just don't
make sense anymore if you have billions of hardware actors (in the
Hewitt sense) following quadrillions of scripts.
-------