Alan J. Flavell

Former Address
Dept of Physics and Astronomy
Glasgow University
Glasgow G12 8QQ
Scotland

There's a rag-bag of things that may be of interest.

NOT my CV:

At University I got to see the EDSAC 2, though not to play with it. We were told that the mercury from the delay lines of the EDSAC 1 was still being kept "in the University Chest" (whatever that meant exactly), in the hope of it increasing in value. I wonder what happened to it subsequently?

In my first fulltime job, at Standard Telecommunications Laboratories, I got to play with the STANTEC ZEBRA. A valve (vacuum-tube) based model, that chucked out lots of heat and needed industrial-strength air conditioning, shared the room with a later version, constructed with germanium transistors that disliked getting even slightly warm: highly unsuitable room-mates. Much of the otherwise "spare" time on the ZEBRA at STL, it played itself extracts of spoken texts and attempted to analyze them - but with what results, I never learned. It was wierd to walk past the unmanned computer room and hear the computer playing snippets of audio tape to itself.

I've found little in the way of description of the STANTEC ZEBRA on the net, though there are several mentions of its name: there was a page at the Rijksuniversiteit Groningen which, after their web site re-organisation can now be found here as De geschiedenis van het RC. This venerable model, although built by STC in the UK, had been designed in the Dutch PTT. In an interesting example of feedback, I subsequently found De ZEBRA, de eerste computer van de RuG there. In the CCS archive, I found a simulator (ZEBRA58.EXE) which I downloaded: I typed in one of my old Simple Code programs (a mortgage interest calculator!) from a faded coding sheet dated 1964, and it ran first time. (Unfortunately the URLs for these archives don't seem stable enough to include them here, but a web search may help.)

One of the features suggestive of the ZEBRA's telephony background was a telephone dial for inputting run-time data. The storage (memory) of the ZEBRA was a magnetic drum, and the trick for efficient programming was to get the instructions and data positioned in an appropriate relative position so that the computer did not have to wait for the drum to rotate to it. There were 2**8=256 tracks of 2**5=32 words, making a total of 2**13=8192 storage locations. Machine code programming sheets had 32 lines, numbered from track relative addresses 0 to 31, to help in laying out the program and data efficiently on the drum. Incidentally, this is the only computer I ever used whose hardware had no default concept of executing instructions one after the other in sequence of storage location addresses - it actually had to be told the address of the next instruction to execute; the other option was to execute the so-called "come-back" instruction out of a register.

I'm told that a core storage version was built later, but I never saw one. When a drum track became unreliable for writing, and would not take arbitrary bit patterns, the trick was to use it as one of the library tracks. Differerent tracks of the library were written to it until one was found that it liked, and then that track was switched read-only.

Later, back at Cambridge, I had not only batch job facilities but also a multi access account (pretty revolutionary for those days) on the TITAN (ATLAS II) machine in the Maths Lab. We accessed it from 10cps teletypes. There, I also learned BCPL, a language for which I have retained a fondness. When Prof. Wilkes discovered that a group of people were playing a competitive game (MOO) on the TITAN system, he initially became very angry at the misuse of resources. However, we heard that some colleagues had depicted the situation to him in a different light, and at a subsequent lecture we heard him praising this work as a practical test of the system's security in that the individual score tables were being updated when each player participated, but without the players being able to interfere with any scores directly.

TITAN worked in some curious units, including the "long second" of 1.28 seconds, and the "short foot" (of plotter output) of 10.24 inches.

When I had the great good fortune to get a job at the Max Planck Institute (for Physics & Astrophysics as it then was) in Munich in 1970, I made the acquaintance not only of the IBM360/91 but also of an elderly valve (vacuum-tube) based computer called the G3, which had been designed and built in-house. This was the place to be in the depths of winter! The machine had some remarkable features, including a hardware stack (called the "Keller" i.e cellar). A later-added peripheral was a device containing a small high-precision white CRT and a colour camera, with a computer-controlled colour filter wheel between. Over many hours, indeed days, of running, this had made a colour cinefilm of simulation of the algorithm that I was to work on, which gave me an excellent introduction to my new job, as well as being a standard feature of our group's presentation at conferences and workshops. It must have been a decade before I saw anything so effective produced by a "more modern" computer: most of them considered a calcomp pen plotter to be high-tech.

The designers had devised a novel scheme for avoiding confusion between I and 1, and between O and 0: the G3 papertape code simply didn't have the letters I and O in its alphabet at all! Words that contained I or O were written with one or zero instead, but, in programming, names were not allowed to begin with a digit, so they had to be written with an initial J and Q respectively, instead.

For our own project, we acquired a PDP8/i (4K 12-bit words!) and a Tektronix 611 storage display. However, the latter was procured on the pretext of being a storage X-Y oscilloscope, in order to avoid it having to be justified to the "computer committee" and bought out of the limited budget for computing - a technique that has stood me in good stead in many subsequent procurements ;-).

When the G3 was finally retired, the hand-crafted "Vorsicht Hochspannung" (caution - high voltage) sign was taken off the wall, and we discovered that it was in fact the back of an original sign with a USA logo and the words "Hier hilft der Marshallplan, zur st?rkung der freien Welt" (the Marshall Plan is helping here to strengthen the free World). The G3, though, was certainly a German production (the programming manual is a fascinating treasury of original German software terminology, much of which has subsequently been overtaken by Americanisms). I believe the "G" stood for G?ttingen, where the institute had been when its forerunner, the G1, had been built.

Original materials ? Copyright 1994 - 2006 A.J.Flavell