22. Januar 2012

Dedicated to my old friends, especially to Bert Forbes, here some rembrances of my Cupertino days back in 1970. Bert was a senior hardware designer at Hewlett-Packard. We all were involved in bringing out the “HP 3000”, as the next model of our minicomputers was named. Bert was CPU design manager. He later left HP and founded Ziatech.
···My group of selected European systems analysts was supposed to learn all about the coming HP 3000. We would be trained in Cupertino and then run a demonstration and support “data center” in HP’s European headquarters, in Geneva. I arrived a bit earlier in Cupertino, I think it was spring of 1970, and set up accomodations, cars etc. for my collegues. They were, see brochure page 15, Marc Brun, Rainer Dern (hardware), Paul Gavarini, John Page, Björn Lindberg (our tallest, to be seen in the picture with the magnetic tape unit), Ray Woodcock, Erich Taschner, myself, plus specialists from the US like Bert Forbes (hardware) and Harlan Andrews. Returning from California late we run the Geneva HP 3000 data center from 1972 to 1974. Please correct me, if I remember wrong. Afterwards I’ve set up the data center at Böblingen, see pictures.

Demonstrating the HP 3000

Minicomputer, a now forgotten race, were square boxes with 16 bit hard and software. We thought in digital three bit numbers, which gives a choice from zero to seven, decimal eight becoming 70, beautiful. IBM’s hexadecimal bytes escalate to A, B, C, D, E and F, far away from us “minis”. To boot a mini you needed to toggle in an initial set of instructions that made the paper tape reader continue the process and load whatever you wanted to run.
···Hewlett-Packard competed against Digital Equipment, who had a broader range of minis up to the large PDP 10, and had a more dynamic instruction set including a hardware register stack. We were static.
···Minis were used for industrial applications, in medicine and military, but we all had fully overlooked any commercial use; a thing for IBM to do, and others like Control Data or Britain’s ICL, on mainframes, magnetic tape and punched card driven.···
···Hewlett-Packard had two computer divisions. The one in Loveland, Colorado, did “calculators” to control Hewlett-Packard’s many instruments, we in Cupertino made minicomputers, more like modular all purpose tech items, real computers (as we thought). In fact Hewlett-Packard missed noticing that they had a PC in their product line, perfect with all-in-a-box computing, ideal for business use. Just like the first “home computers” HP’s calculators had no operating system, but run Basic natively. You turned them an, and there was either an application that started on top of Basic or Basic. Basic commands like “goto” or “for” etc. where printed along the keys on the keyboard as typing shortcuts, I think. While Ataris were toys, HP calculators were serious stuff – and seriously expensive. Nobody even dreamt of calling them “personal computers”.
···HP 2116. At the “real” HP computer division in Cupertino Page Mill Road we had operating systems controlling the hardware. Models were HP 2116, later 2115 and the “budget” 2114, see ad: “Is it possible to get a really good computer for less than $10K?”, 4k×16 bit.
   At first a “basic input output system” (BIOS) run a single stream of code, the program had to loop while waiting for a typed input or a character from the punch tape reader. SFS, skip if flag set, was probably the command excercised most of the time. (See an ad from 1970, an ad for a disc drive fom 1971). Then in 1969 a hardware interrupt system allowed the program to progress while waiting for the next character from the reader, a big step to speed it all up. A interrupt was just a hardware-forced jump-subroutine to a specific memory location. With the interrupts turned on, a “real time operating” system became possible. My group of systems analysts, then in Milan, was assigned to debug it, great intellectual fun.
···Interrupts allowed another feature: You could make one minicomputer work for more than one user. If you knew from which teletype the characters came in, you could assemble them to input data or commands strings in different bags for different users – up to 32 in those days. Initially this was done by a beautiful software driver shifting bits until it got to the correct user, later by a special board. As playing withe the operating system was reserved for the real experts (nowadys called “administrator”), the users commanded in Basic, hopefully not able to crash the system. Basic was general knowledge, and Basic was an interpreter, i. e. it let you know programming errors right away, not only after a lenghthy compiling process translating into machine code. Compilation (Fortran) took up to four passes with intermediate paper tapes on a 2114.
···Incidentally, my Basic interpreter, 55 kbyte of code, copyright Microsoft 1985, still runs on todays PCs!
···Back in the late 1960s and early 70s with Basic time sharing users in schools or administration typing along on their teletypes got connected directly or via 110 bit per second telephone modems to the central minicomputer system, initally a single computer with 64 kwords (à 16 bits, equal 32 kbyte), its adressable maximum, plus a disk. See ads for it from 1970 here and here.
···The HP 3000 was supposed to be Hewlett-Packard’s next computer model. Developement had progressed on a 32-bit computer named Omega, stack oriented, very dynamic; but to top management this seemed too big a step. So another, more humble project with just 16 bits, later called Alpha, was chosen for further life. It had been started by started by Mike Green, Alan Hewer and Bert Forbes, who comments today: “I don’t recall the original codename – HP3000 was chosen much later. It got renamed the Alpha and suddenly had many people from the Omega team trying to help with it. The expectations for what it needed to do increased by an order of magnitude to include batch, real-time and time-sharing. Quite a load for a sixteen bit computer of the day.” (I used to have a copy of the Omega specs. Probably can’t find them any more. Bert says there is a copy at the Computer Museum archives in Mountain View, CA.)
···The Alpha was very much state of the art, even bejond, and had relative addressing via pointers for the code, so various users would be running their own instance. The stack addressing allowed subroutines to be re-enterable. Data addressing in the planned Omega had been relative as well, but remained absolute in the Alpha, other than the stack. “And we made the classic mistake of ‘who'll ever need more than 64k (words)?’ That mistake has been repeated by every generation of computer designers from Bill Gates (640k bytes) to IP4”, so Bert Forbes today.
···I remember the Alpha had relativly large boards, horizontally stacked, and at one time you had to insert an indiarubber between two baords to make the machine run. Probably a connection was broken, unless the board was purpously warped. Bert Forbes knew how.
···The software was far behind schedule. It was supposed to do all the next advances from time share Basic, but in all directions at the same time: not just Basic, simultaneously more languages like Fortran, an assembler, and later even Cobol “for educational use”, not just 32 users, 64, etc. The system should be so fast that it would finish your typed in commands outrunning your fingers (like Google completes your searches while you type in). And so on, dreams, and as I like to say: A unborn child is everybody’s wish, both boy and girl.
···Finally the HP 3000 was due to be introduced at the 1971 Fall Joint Computer Conference in Anaheim, CA. My European group at that time was already working in Cupertino – see page 14 of www.Joern.com/FritzFolio.pdf – and got assigned to write the demonstration program. No operating system software was yet running on the machine, and if at all, it was slow, slow, slow with its obesity. However the time sharing round robin software for the terminals was up. So I suggested a “advertizing” demo. A screen page had 25 lines, if I rememer well, so we had marketing produce texts with 25 lines à 80 characters, 27 of them for A to Z. You typed in F, and some promotional text with F in the headine came up, right out of memory (we had no mass storage running either, but plenty of core memory), etc. The prototype HP 3000 was preloaded with 675 (27×25) punched cards into memory, and off it went to the show. The demo was super fast, outperforming any future reality.
···In Europe we never oversold the early HP 3000. With its RPG compiler it made a good competitor to IBM’s System/3, being able to run multiple threads for multiple users. So, back in Geneva with my group, we remained unshaken by the sales troubles in the US.

Permalink to here:

Meeting in June 2017 http://blogabissl.blogspot.com/2017/06/california-revisited.html#friends
Time Table (hopefully correct)
late 1970 Data Center Milano
1971 and 1972 – Fritz Jörn in Cupertino, initially alone preparing for the teamn to come
July 1972 to November 1972 ― the European HP 3000 team is trained in Cupertino
November 1972 to February 1974 ― European HP 3000 support out of Geneva
···John Page left August 1973 and went back to England
1975 ― HP 3000 support from Böblingen

HP: “A decade of large scale innovation” – In my hindsight opinion HP’s separation and rivalry of computers (in California)  and calulators (in Colorado) was tragic. HP never noticed that it had the perfect PC at hand: the calculator with its built-in Basic programming. If they just would have added Visicalc

For some of my pictures of the early HP 3000 please see
including a German brochure and price list from 1980.
The brochure and price list sepatately as PDF here.
The HP 3000 demo data center in Geneva, a brochure
See http://en.wikipedia.org/wiki/HP_3000
History (by Bob Green): http://www.robelle.com/library/smugbook/classic.html
More history (by Christopher Edler): http://www.3k.com/index_papers_hp3000_history.html
Promotional brochure, 1971, “System Description”
HP “Museum”: http://www.3k.com/index_papers_hp3000_history.html – The picture up on the right is from there. A while the HP 3000 came in different colors, the panels could be individually designed, but this delayed sales decisions beyond reason.

As you see here, 4.9 MByte of disc memory cost $ 9975 back in 1971. My German price list of July 1980 shows a “7906M” 20 MB “Master” for 34550 DM plus HPIB »Adapter« for 2310 DM, a total of 36860 DM (€ 18846, appr. $ 23606), without tax.

A comment by Bob Strand to the picture with the many terminals (http://picasaweb.google.com/Fritz.Joern/HP3000#5700834382860552162): That picture with the multiple 3000 terminals was during the testing we did in Cupertino for Makro stores in The Netherlands. I moved to Amsterdam in December of 1975 and my job there was to consult with the Makro guys who were to write the software for the dual 3000 systems to run their stores in Holland, Belgium, The UK and Spain. This was to replace their current system that ran on 2100’s that had been developed by Dave Mackie [HP2100: see http://en.wikipedia.org/wiki/HP_2100 for good histrory. fj] It was soon apparent to me that they did not have the expertise to do the job. After months of struggling, things came to a head in a big meeting in Utrecht. During the first half of the meeting, they spent the entire time blaming their lack of progress entirely on me. If I had been a better consultant, they would have been further along than they were. After an hour and a half like this, all of us technical people were excused from the meeting and it continued with only the HP and Makro execs - Jan Schapers country manager from The Netherlands and the HP Europe manager (Dick Anderson?) who had come up from Geneva.
···After another hour or two the management meeting ended. Makro's conclusion after this part of the meeting was to turn the entire technical project over to me - the totally incompetent person from the first half of the meeting. They had used about five people for six months to get nowhere. I started from scratch and wrote the whole thing in about four months, all in SPL. (I actually still have the line printer listing of the program on green and white striped paper. Many hundreds of pages.) Once finished, HP sent a manager for me from Cupertino to take all the credit. I left and went to Tandem in February of 1977. I had started at HP in October of 1972 as an instructor in the HP-3000 training department. Met your [Fritz Jörn’s] team there and came to Geneva in the summer of 1973 with Unanski to teach the new release of MPE. You hosted the really nice graduation dinner for that class on my 31st birthday - July 18. That was my first ever trip to Europe.

Keine Kommentare: