Jecel --

Remember that BASIC on the Altair didn't do much. I don't think that a
Smalltalk type system would have to be much larger if it "didn't do much".
Peter Deutsch's original interactive PDP-1 LISP from the early sixties did
do quite a bit and it was implemented in about 2000 instructions on a 4K
(18 bit word) PDP-1. Smalltalk-72 did quite a bit and it ran quite well in
about 16kb + the display memory. It would be interesting to see how well
the "page long" interpreter for ST-72* would do if all that were added were
a few auxillary classes.

Of course, old PARC hands would also point out that the Altair was quite
irrelevant since the Alto at PARC started running two years earlier in
1973, and there were quite a few of them by 1975. All personal computers
today are like the Alto, none are like the Altair. The hobbiest PC movement
was kind of a red herring and dead end (and one could say that 8-bit micros
and most of the software that was put on them led several generations
astray from better ways to approach personal computing. Those bad defacto
standards are still holding back progress).

* there's a version of it sketched in the "Early History of Smalltalk"
chapter (in an appendix) in the HOPL II book.

Cheers,

Alan

-------

At 3:05 PM -0800 5/12/00, Jecel Assumpcao Jr wrote:
>On Wed, 10 May 2000, Peter Crowther wrote:
>> Alan?  Dan?  Fancy building Squeak on a machine that only has one
>> instruction? :-)
>
>Actually, I am doing just that. The internal architecture of the
>processor only has the conditional move instruction while the user
>visible architecture is the bytecodes (Self bytecodes rather than
>Squeak ones, but that is changeable...). Ok, so I cheat - there are
>various memory mapped functional units (so when you move data to two
>addresses their sum magically appears at a third address).
>
>This thread has shown me that people on this list like to stray off
>topic and they like to remember the old days. So I thought I might
>share a bit of idle speculation I sometimes enjoy:
>
>Imagine that one day you wake up and find you have been transported
>back in time to January of 1975. You still remember most events as they
>originally happened up to today and, better yet, find your trusty
>laptop is still with you. On it you have the latest release of Squeak
>(who would be caught without it?) and lots of junk, including a very
>nice Altair emulator.
>
>Now here is my question - using all this, can you create something that
>will run on the 4KB Altair that will be so superior to the future
>Micro-Soft Basic that you will change history and take computing in a
>better direction? Remember, both the language/system itself *and* one
>reasonable application must run in that 4KB.
>
>If you thought Forth, then you were close but not quite there. Forth
>was, in fact, available back then and could have run on that machine.
>But it wasn't user friendly enough to spread beyond a small group.
>
>If you thought tinySqueak, that is what I am trying to imagine. But all
>of the efforts to run Smalltalk on 8 bit processors that I am aware of
>needed 64KB or more (TinyTalk by Larry Tesler and Kim McCall,
>Smalltalk-PC by Christopher Macie). Smalltalk-72 was amazingly compact,
>but I am not sure even it would fit.
>
>What got me thinking about all this was an evaluation I did a while ago
>to see if the ARM7500 would make a reasonable Squeak machine (not
>quite). So I began to wonder if the stuff we do has a kind of lower
>bound. Must early computing really start with Basic and friends and
>must better things wait for the arrival of "real machines"?
>
>Cheers,
>-- Jecel