And if you look at the photo of the front panel, you can see the "AUTO RSTRT" switch that's mentioned in the article - it's the third from the left on the top panel:
We had one of these refrigerator-sized boxes in our machine room in the early 80's to connect our DEC-20 to the Arpanet. A few years later, it was replaced by a Cisco router that was a fraction of its size.
> The IMP software and the ARPA network communications protocol running on the IMPs was discussed in RFC 1 (https://datatracker.ietf.org/doc/html/rfc1), the first of a series of standardization documents published by what later became the Internet Engineering Task Force.
It was also pretty clear that the IMP itself, and BBN as its vendor, was kind of at the center of this whole thing, where RFCs were always implicitly for internal discussion between these people who were also always in external discussion with BBN, recording what BBN said and what changes to the IMP were coming that would allow for/enable changes to planned protocols and software.
You might say that the IETF was really just a user group for "amateur" BBN IMP enthusiasts. :) I say "amateur", because BBN's real paying customers, as a DARPA contractor, would have been government, military, and telcos. Thus why none of those were participants in the "Network Working Group" — they were getting white-glove service directly from BBN!
If anyone knows exactly when DARPA itself (and US gov/mil installations as a consequence) began participating in the IETF / switched their own internal networks to using IETF-developed standards, that'd likely be a fascinating little point of history to learn about.
When I worked at BBN in the late 90s, around the time they were bought by GTE, they had a warehouse of old equipment, and I noticed they had a couple of IMPs. I really hope those ended up in a museum.
