[COFF] GE-635 GECOS?

[Douglas McIlroy via COFF]

Clem wrote

> Check out
> https://www.multicians.org/ge635.html

> From what I can tell GCOS has been lost to time

The Multicians timeline shows Bell Labs joining the project after GE
was chosen as the vendor. This may be the date of some formal
agreement. However, the Labs was in on the final confirmation of GE.

IBM had been ruled out over the issue of virtual memory. GE was happy
to offer a VM variant of the 635. In contrast, IBM's chief architect,
Gene Amdahl, adamantly maintained that VM was unacceptably
inefficient. When it became clear that GE was winning, IBM revealed
Gerrit Blaauw's 360 model 67 with VM, which had been kept under wraps.
The option was considered, but found to offer no advantage from a
programming standpoint. Nevertheless, time-sharing on the IBM machine
made it into the field well before Multics; MTS, the Michigan Terminal
System, went live in 1967. the same year that GE delivered the 645.

A completely unsung project at Bell Labs during the Multics era was
Joel Sturman's GUTS, GE User Time Sharing on the 635. Joel built this
without any specific kernel support. I'm sorry I don't know more about
it, in particular how it did scheduling.

Way off topic, but apropos of the 360 vs contemporary 36-bit machines,
a member of the design team for the Apollo mission once told me that
it was fortunate they were using IBM 700/7000 equipment rather than
the new-fangled 360s. 36-bit floating-point precision was just enough
to calculate trajectories to the moon without resorting to slow
software double precision. The 32-bit floating point of the 360 would
have forced the slow alternative. I suspect also that hex rounding
would have exacerbated the difference.

Doug