You make good points. I typed ASAP. There are a lot of
possible "quibbles" with what I wrote.
You are correct about the 'vision' thing -- Tom, Jr.
did a very gutsy thing pushing out System 360. He
essentially bet the company. By then IBM was
well into computing, e.g., the 7094 based on
transistors instead of tubes, but 360 was
at least two giant steps more.
But I do remember
that somehow about then DEC was able to do
the PDP 10, nice system -- hmm. Dartmouth
was able to do the system that GE used to
sell time-sharing, etc. MIT did Multics.
Net, others were able to write OS software without
betting a Fortune 10 or so company on the work.
You are fully correct about the anti-trust suit -- in
a sense it made IBM 'gun shy' and timid for a long time.
As I recall, long IBM sold the main chips
used by Cisco and Juniper. I can guess that
IBM gets patent licensing revenue for
disk head technology, and maybe the heads
themselves. But the big bucks are in
selling the drives and the
subsystems.
I forgot about the sale of Maxtor to Seagate.
I guess my broad point was that IBM was essentially
always a 'marketing company' run by successful
mainframe salesmen. E.g., once, the day after my wedding,
I got an offer from the IBM Chicago branch office, and the
head there gave me that description of IBM. They wanted
me to hold the hands of the oil refining customers using
linear programming to make decisions on what inputs to take
and what outputs to make -- big bucks in that, still. They had sold a 360/85 and likely wanted to have sold more. Gee, I might have gotten to work with Ralph Gomory, later head of IBM Research.
I guess part of my 'broad point' was that they kept
throwing off their plates little opportunities like
Intel, Microsoft, Cisco, Oracle, Seagate, EMC, operating the whole
Internet (can you believe that!), Yahoo (IBM had
Prodigy, good idea, bad execution), OS/2 (ahead of
anything from Microsoft until Windows NT or 2000),
Netscape (IBM had a decent Web browser early on),
Web servers ("ah, what's to do with an Web server;
trivial, right?" -- build one easy to program and to serve 10,000 pages a second and then tell me that).
PL/I? Yup! It was done by a committee headed by
George Radin in about 1964. I used PL/I and CP67/CMS to
do the first computer based scheduling of the fleet
at FedEx. Nicely enough I was paid well
to learn PL/I by the US DoD at the
JHU/APL for some work on passive sonar and the
FFT. Once I tried to talk to Radin about
operating systems, and he said, "Three times
in my career I tried to help IBM in operating
systems, and three times I broke my pick trying.".
I remember the sale to Hitachi but didn't know that
the IBM drives sucked. I've heard only good things
about Hitachi drives; sorry I didn't mention them --
another quibble.
IBM had object oriented programming in
microcode as part of 'Future Systems'
in 1980 or was it 1970?
But, they were a 'marketing company' selling to
'good IBM customers'.
> Although that also seems to be happening in services.
I can believe that. Long IBM had their pick of
the job pool something like Google seems to today;
likely no more.
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.
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.
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).
You make good points. I typed ASAP. There are a lot of possible "quibbles" with what I wrote.
You are correct about the 'vision' thing -- Tom, Jr. did a very gutsy thing pushing out System 360. He essentially bet the company. By then IBM was well into computing, e.g., the 7094 based on transistors instead of tubes, but 360 was at least two giant steps more.
But I do remember that somehow about then DEC was able to do the PDP 10, nice system -- hmm. Dartmouth was able to do the system that GE used to sell time-sharing, etc. MIT did Multics. Net, others were able to write OS software without betting a Fortune 10 or so company on the work.
You are fully correct about the anti-trust suit -- in a sense it made IBM 'gun shy' and timid for a long time.
As I recall, long IBM sold the main chips used by Cisco and Juniper. I can guess that IBM gets patent licensing revenue for disk head technology, and maybe the heads themselves. But the big bucks are in selling the drives and the subsystems.
I forgot about the sale of Maxtor to Seagate.
I guess my broad point was that IBM was essentially always a 'marketing company' run by successful mainframe salesmen. E.g., once, the day after my wedding, I got an offer from the IBM Chicago branch office, and the head there gave me that description of IBM. They wanted me to hold the hands of the oil refining customers using linear programming to make decisions on what inputs to take and what outputs to make -- big bucks in that, still. They had sold a 360/85 and likely wanted to have sold more. Gee, I might have gotten to work with Ralph Gomory, later head of IBM Research.
I guess part of my 'broad point' was that they kept throwing off their plates little opportunities like Intel, Microsoft, Cisco, Oracle, Seagate, EMC, operating the whole Internet (can you believe that!), Yahoo (IBM had Prodigy, good idea, bad execution), OS/2 (ahead of anything from Microsoft until Windows NT or 2000), Netscape (IBM had a decent Web browser early on), Web servers ("ah, what's to do with an Web server; trivial, right?" -- build one easy to program and to serve 10,000 pages a second and then tell me that).
PL/I? Yup! It was done by a committee headed by George Radin in about 1964. I used PL/I and CP67/CMS to do the first computer based scheduling of the fleet at FedEx. Nicely enough I was paid well to learn PL/I by the US DoD at the JHU/APL for some work on passive sonar and the FFT. Once I tried to talk to Radin about operating systems, and he said, "Three times in my career I tried to help IBM in operating systems, and three times I broke my pick trying.".
I remember the sale to Hitachi but didn't know that the IBM drives sucked. I've heard only good things about Hitachi drives; sorry I didn't mention them -- another quibble.
IBM had object oriented programming in microcode as part of 'Future Systems' in 1980 or was it 1970?
But, they were a 'marketing company' selling to 'good IBM customers'.
> Although that also seems to be happening in services.
I can believe that. Long IBM had their pick of the job pool something like Google seems to today; likely no more.
So be it.