Mercury delay line was so cool. In one long tube filled with mercury, a transmitter on one end of the tube, and a received on the other. Using timing, you could tap about 1000 bits before the first of the 1000 would propagate to the receiver at the end of the tube. The “delay line” concept likely stems from the time it takes for the first tap to travel to the end of the tube. Once the tap travels through the mercury and hits the receiver, the first bit to arrive would need to be regenerated. Those 1000 bits would be all you got from that tube. The pulses were ultimately encoded so the timing wouldn’t be lost if you had a stream of zeros.

21:47 Licklider became head of ARPA, which sponsored this project called the “ARPAnet”, which later turned into the “Internet”.

37:33 Pages are fixed-length pieces of memory, segments are variable-length. I think MULTICS had both. Also note that its segmentation scheme operated within a flat address space. Unlike the later horrible kludge that was the x86 architecture, just for example.

44:39 All these other influential systems had one thing in common: unlike Unix, none of them was portable. They were each developed for just one particular architecture. That’s probably why only Unix is remembered today.

22:37 Flexowriters were modified IBM Selectric typewriters, as I recall. They had added to them the ability to punch and read paper tape, as well as connect to a computer. A lot of systems used the ASR33 or KSR33 Teletype devices as terminals instead. These connected to computers via a serial 20mA current loop interface. Not sure how the Flexowriters connected.

15:51 John McCarthy was a total IBM man. When Stanford wanted to buy a Burroughs machine, IBM bribed them by giving them one of its units for free, plus a hefty cash grant. So Stanford took the IBM machine, and used part of the cash grant to get the Burroughs one as well. The students preferred the Burroughs machine, because of its faster turnaround time. But McCarthy saw to it that that feature was crippled, until finally people stopped using the Burroughs any more. ed-thelen.org/comp-hist/B5000-AlgolRWaychoff.html

45:25 Paging hardware had to be custom-built for MULTICS and also TENEX. That kind of thing was simply not common at the time.

Where are the sources? Most important thing, the bibliography!?

43:00 Here is a good, detailed presentation on MULTICS, from someone with first-hand experience with it: kzfaq.info/get/bejne/Zs2qndV3ntiagXU.html . Honeywell put out a brochure in 1982, available here bitsavers.trailing-edge.com/pdf/honeywell/multics/DL92-01_multicsBrochure_1982.pdf . And if you compare that to any other OS available in 1982, you see how advanced MULTICS was, even into the 1980s. According to Steve Webber, Honeywell had no idea what to do with the OS. They deliberately priced it at an eye-watering seven-figure price. Yet enough customers still bought it to make it profitable. And they were very loyal to it.

15:15 Presumably that is this one kzfaq.info/get/bejne/oNOCeJihvazNYGg.html .

Analog computer cant be slow... There must be a design flaw in it or some other reason to give up on it to use as a flight simulator 😮

Poor guy is massively uninformed as demonstrated by his complete unawareness of the Digital PDP-6 and PDP-10 family KA, KI, KL-10 processes providing commercially available time sharing starting in the 1960s. The major time sharing service CompuServe was based on PDP-10 mainframes running “the monitor” because the os never had a name for years. Eventually named TOPS-10. BBN added a paging box to create TENEX. PDP-10 time sharing user loads ranged from dozens up to just under a hundred users with uptimes of six months between re-boots. PDP-10s at national labs provided terminal front ends for clusters of CDC 6600/7600 mainframes, implemented the Ramada Inns inline reservation system. Bill Gates and Paul Allen learned to program on a Ten terminal.

i couldnt make it past 11 minutes of this video. The speaker cant make it through a single sentence without saying "um" or "uh" four of five times.