8 Apr
2004
8 Apr
'04
5:10 a.m.
>>>> "Andy" == Andy Holt
<andyh(a)andyh-rayleigh.freeserve.co.uk> writes:
> The Cybers weren't just the world's best
compute engines by a
> large margin. They also have excellent I/O performance.
Andy> Yes, but
...
> Show me another 1965 vintage computer that can
support timesharing
> with 600 active users and excellent interactive response
> time... (I'm talking about the PLATO system here -- which ran 600
> users on a 6400, not even a 6600, though the I/O is the same for
> either.)
Andy> That was a special-purpose TS application - comparable, say, to
Andy> airline booking (which I understood that both IBM and Honeywell
Andy> (and Univac?) managed to do reasonably competently).
Not really. The amount of interaction was 1-2 orders of magnitude
higher. Airline reservation systems didn't require per-keystroke
processing.
But you're right, it certainly is possible to build crummy timesharing
systems on any architecture. I don't recognize "Respond" but I
remember "telex" on CDC NOS. I also remember (vaguely) "RAX" on IBM
360s...
I think we've been arguing the question "did IBM do more or better
with the 360 than CDC with the 6600". You can argue that it did, and
that some of this comes from having more stuff. (Then again, CDC had
at least three high level programming languages and probably more than
that...)
Perhaps more interesting would be who did better or more on a
productivity basis -- stuff per person or stuff per $.
paul