5 ms·

"MIT did Multics" And it was a fairly ugly, long, and drawn out second system syndrome experience; rather famously Bell Labs dropped out of the project but the

by hga 12y ago

"MIT did Multics" And it was a fairly ugly, long, and drawn out second system syndrome experience; rather famously Bell Labs dropped out of the project but the experience inspired UNIX(TM), which is Multics with some vital parts missing. Delivered much later than planned, but it was a quality system. It very possibly was the only system of the ones you mentioned that had scope on the order of OS/360 and TSS/360 (5 and 1 man millennia) ... but it wasn't done with the commercial pressure on those IBM OSes, and of course embraced virtual memory vs. disdained it. Multics and the PDP-10 were both crippled by 1 MiB address space sizes (1/2 of 36 bits, word addressed, the total for the PDP-6/10/Decsystem-20, per segment for Multics, which was much less limiting, mostly a nightmare for big data sets). I'd quibble OS/2 was also a failure of execution, driven by marketing. IBM told IT managers that the PC/AT was the last PC model they'd have to buy for a long time, but the 286's protected mode was horribly misconceived (changing a 64 KiB segment incurred a terrible performance hit). Therefore early OS/2 had to run well on it ... but really couldn't all that well. Whereas Windows 3.0 hit a lot of niches very well, and the rest is history. Somehow I don't see Cisco, very much not part of their DNA ... but then again, that's your whole point. Future Systems would have been circa 1970, it was the ambitious 360 follow-on (vs. System/370 which was 360s with ICs and DRAM). As I understand it, that group eventually gifted us with the very advanced AS/400 et. al. Like Multics, every file is part of the address space, it's very neat and worth studying. Anyway, thanks a lot for the insights and stories you've shared. I got my start with the IBM/1130, but then it was UNIX(TM), watched but didn't really partake of the by then doomed Multics (Honeywell was horribly managed, blamed a project failure on the decision to microcode the machine and then tried to compete with 1 MPS async processors through the '80s), PDP-10s and Lisp Machines, followed by an unending sequence of UNIX(TM) and Unix alikes. Bleah, when you know we can do much better. But it didn't have to be that way. IBM certainly had a chance to win my heart and mind with their systems in my home town's college (the 1130 and a 370/115), before I got exposed to better systems when I left.

graycat 12y ago
The story went, some guys at Honeywell on Multics went to management and said, "We believe that we can bring up Multics on a super mini computer, sell it, and make money." and management said, "We don't believe you can bring up a Multics on a super mini computer; if you did, it wouldn't sell; even if it did, you wouldn't make any money.". So, those guys did Prime. The OS was written in a slightly tweaked Fortran. I ran two of them and did my dissertation computing on one of them. For the first one, we were doing some DoD analytical work on TSO, and our two programmers were spending $80 K a year. We got a Prime for $120 K, and our computer usage went through the roof. We just copied over our 500 K TSO Fortran programs, compiled, and ran -- ran fine. That Prime was just for our group of 40, but soon the company of 300 wanted one, and I served on the selection committee. For the second one, I got that for a B-school as a prof. There the Prime made the central computer group's Amdahl 470/V6 look silly, and I served on the committee to select a new university CIO. At one time I had a summer job, and at first I was to program an 1130, but later the job was for me to design and build some DC power supplies to power some IBM tape drives IBM had given away to a research lab.
- hga 12y ago
  Software Arts, as in Visicalc, used Prime computers. I see from a biography of Dan Bricklin, the less technical of the two, that: Prior to forming Software Arts, he had been a market researcher for Prime Computer Inc., a senior systems programmer for FasFax Corporation, and a senior software engineer for Digital Equipment Corporation. At Digital, he was project leader of the WPS-8 word processing software, where he helped to specify and develop one of the first standalone word processing systems. And the more technical partner Bob Frankston was an MIT type. I remember one Software Arts employee, friend of one or more, or maybe Bob's youngest brother, who was a friend of mine, mention that among other things they appreciated what the system adopted from Multics. Being bit-sliced, the Prime micro-architecture you mention was microcoded, which of course would fit with their following the example of the successful System 360, many models of which had to be microcoded because the logic family they used pretty much had only one speed, they made micro-architectures narrower, down to 8 bits as I recall, for the slower machines, and the 2? fastest had none. Honeywell's rejection of microcoding helped made Multics and GECOS systems terribly uncompetitive, at least by the time Visicalc was being developed, although the macro-architecture allowed you to easily hook up 6 CPUs in one system (and 8 with a horrible kludge, as I recall).

by hga 12y ago

graycat 12y ago

The story went, some guys at Honeywell on Multics went to management and said, "We believe that we can bring up Multics on a super mini computer, sell it, and make money." and management said, "We don't believe you can bring up a Multics on a super mini computer; if you did, it wouldn't sell; even if it did, you wouldn't make any money.". So, those guys did Prime. The OS was written in a slightly tweaked Fortran. I ran two of them and did my dissertation computing on one of them. For the first one, we were doing some DoD analytical work on TSO, and our two programmers were spending $80 K a year. We got a Prime for $120 K, and our computer usage went through the roof. We just copied over our 500 K TSO Fortran programs, compiled, and ran -- ran fine. That Prime was just for our group of 40, but soon the company of 300 wanted one, and I served on the selection committee. For the second one, I got that for a B-school as a prof. There the Prime made the central computer group's Amdahl 470/V6 look silly, and I served on the committee to select a new university CIO. At one time I had a summer job, and at first I was to program an 1130, but later the job was for me to design and build some DC power supplies to power some IBM tape drives IBM had given away to a research lab.

hga 12y ago
Software Arts, as in Visicalc, used Prime computers. I see from a biography of Dan Bricklin, the less technical of the two, that: Prior to forming Software Arts, he had been a market researcher for Prime Computer Inc., a senior systems programmer for FasFax Corporation, and a senior software engineer for Digital Equipment Corporation. At Digital, he was project leader of the WPS-8 word processing software, where he helped to specify and develop one of the first standalone word processing systems. And the more technical partner Bob Frankston was an MIT type. I remember one Software Arts employee, friend of one or more, or maybe Bob's youngest brother, who was a friend of mine, mention that among other things they appreciated what the system adopted from Multics. Being bit-sliced, the Prime micro-architecture you mention was microcoded, which of course would fit with their following the example of the successful System 360, many models of which had to be microcoded because the logic family they used pretty much had only one speed, they made micro-architectures narrower, down to 8 bits as I recall, for the slower machines, and the 2? fastest had none. Honeywell's rejection of microcoding helped made Multics and GECOS systems terribly uncompetitive, at least by the time Visicalc was being developed, although the macro-architecture allowed you to easily hook up 6 CPUs in one system (and 8 with a horrible kludge, as I recall).