9 Jun
2006
9 Jun
'06
4:37 p.m.
On 6/9/2006 at 11:29 AM Billy Pettit wrote:
4. The most exotic peripheral I saw was an 18"
wide tape unit that used a
helical scan head bar. It was being developed by the Government Systems
Division. It made so much noise that you had to be in another room when
it
was running.
The infamous "scroll" drive? That was supposed to make it to the Star
platform too.
Exotic? I rememer that Jim Thornton (and later Niel Lincoln) were off on
an EBAM crusade for awhile. The last I saw of an EBAM unit was sitting in
a hallway at ADL. Presumably to be scrapped, I don't know.
The Star 1Bs were incredibly unreliable, particularly given their speed.
You'd take a tape with the OS to be built on it and it'd take most of the
night--if the system didn't die in the meantime. Often, it was just easier
to cage some time at LLL and do it there. I didn't have any contact with
the 65, but the folks who came to Sunnyvale from Canada had fond memories
of it. I'd proposed writing an emulator for the Cyber 70 to at least get
compilations and builds done, but that got nowhere--I suspect it would've
been faster than the 1Bs.
The Star-100 had a lot of problems, both with the archiitecture and
personnel--the instruction set was huge (if you could think of an
instruction, it was probably there). Running everything through the vector
units, including scalar ops (can you say "start up latency"?). There was
an insufficient understanding of how to manage virtual memory; it wasn't
until fairly late in the project that working-set paging got fully
debugged.
The division of effort between Arden Hills and Canada, and then later,
Sunnyvale made for bad communication and attitudes. I came into Star when
it was given to Nix Frazier and he brought some of the old Sunnyvale SSD
personnel to the project. The attitude in the original software group was
unbelievable--like a bunch of civil service lifers. The OS being used by
pretty much the only customer didn't originate at CDC, but at LLL, was
written in a dialect of LRLTran (IMPL) and contained very little vector
code--and what code that was generated was horrible.
A lot of the architectural problems were addressed in the Cyber 203, but I
think it was too little, too late.
Most of the sharper folks at Sunnyvale moved to outside jobs in the late
70's, taking most everthing that they learned with them. I didn't hear
from Neil Lincoln again until about 1983, when he was
off on the ETA-10
(nee GF-10). The new digs in St. Paul were pretty cool, but the
same old
personnel and mentality was there. I'd proposed to Neil that doing a port
of Unix might be a good shortcut to getting an OS going, but that was
overruled. So a couple of buddies and I started a firm in Santa Clara
whose sole purpose was to do an optimizing FORTRAN for the ETA-10. I left
about 1987 to go out on my own--that was my last contact with really big
iron, though I did get a call sometime around then from Neil wondering if I
could do a port of Unix to the ETA machine in 6 months...
Cheers,
Chuck