[ih] ARPAnet and nuclear survivability

[Jack Haverty]

On Wed, 2009-09-09 at 13:46 +0200, Matthias Bärwolff wrote:
> The "BBN versus graduate students at host sites" issue 

This was a real issue.  I remember clearly the consternation on the DEC
field service guy's face when he came to MIT to service the AI Lab
PDP-10 and discovered that the instruction set had been changed.  It was
helpful to have a "rotate" instruction that went "the other way" for use
by the Chess program, so someone (Tom Knight?  Jack Holloway?) simply
added the hardware to make an unused opcode do something useful.

BBN was effectively an MIT spinoff, just up the road from MIT, and
heavily populated by ex-graduate students.  The IMPs were being
delivered to MIT and other similar universities.

I'm surprised the cabinet wasn't welded shut.

Later on, when the IMPs were out in military sites, we had one machine
that was consistently unreliable - rebooting every few days for no
apparent reason.  After field service had replaced everything except the
cabinet, a field service guy was assigned to sit in the computer room
and wait for the IMP to reboot to see if he could learn anything.

After a while, a door slammed somewhere, and a older colonel stormed
into the computer room, cursing and complaining, went to the IMP, opened
the cabinet door, opened the top of the computer itself, reached inside,
and pushed the internal reset switch on the main board.

Turns out that he knew that when comms didn't work, he knew from
experience that all you had to do was recycle the box, so when his email
didn't work...he'd reboot the IMP.

Must have been a graduate student before Arpanet time.

I think the cabinets were all locked after that.  Frank's concern was
well-founded.

/Jack Haverty
Point Arena, CA
ex-MIT (1966-1978)
ex-BBN (1978-1990)