|Resurrection Home||Previous issue||Next issue||View Original Cover|
The Journal of the Computer Conservation Society
|Corrections and Clarifications|
|Queries and Notes|
|Alan Turing and Oxford||Jonathan P. Bowan|
|A Life of Calculation (Part 2)||Ron Weeden|
|Another Long-Undiscovered Software Error||Dik Leatherdale|
|50 Years Ago .... From the Pages of Computer Weekly||Brian Aldous|
|Committee of the Society|
|Aims and Objectives|
Elliott 803, 903 & 920M — Terry Froggatt
Whilst we still have no circuit diagrams for the 920M, we probably have too many for the 903. I have the circuits which came with my two machines and from the Users Association plus missing sections scrounged from Elliotts Rochester in the late 1970s. Andrew Herbert has scanned the circuits which came with Don Hunter’s two machines (now both in Cambridge) and Oliver Harlow’s 903 at TNMoC. Some are A4 (or folded A3) in A4 ring binders; others are “post” size (15” by 19”) held flat in large ring binders. I’ve spent some time recently looking for variations in these sets of circuits. These include extra wiring on the A-FR card (and a backplane change) to support extra store, an extra bistable on the A-GF card to disable initial instructions, a timing capacitor change on the A-FQ card, and extra pull-up resistors on the A-FA main register cards. These variations were previously known, and we have (for example) added the missing A-FA resistors to the TNMoC 903. But new to me was an extra 2.2K pull-up resistor R9 on the A-FI cards. And sure enough, when I pulled the A-FI cards on my two machines, one had R9 and one did not. I’ll be checking the other 903s in due course.
In Resurrection 80, I reported that an error had been found (and corrected) in the ATN function of the 903 “Workshop” calculator program, written in the Department of Transportation and Environmental Planning at Birmingham University, and issued in May 1970 in Marconi Elliott house-style. I noted then that a bonus from this investigation was the discovery that Workshop uses a four-word software floating-point format interpreter, which could be a useful addition to the two-word and three-word “QF” formats used by 903 Algol. I rather hoped that this would implement four-word division using a loop-free handful of hardware divisions, along with some multiplications and subtractions to recover the remainders. I’ve now looked into the code, and find that it uses no hardware division. Rather disappointingly, it uses a loop which calculates the 51 mantissa bits, one at a time, by shifting and subtraction.
I’ve added an extra facility to the PC-hosted 2-Pass SIR which I announced in the previous issue of Resurrection. The usual SIR notation “^X” causes the subsequent words to be placed from address X onwards (where X is an absolute address or located label), and I’ve added “,X” which only changes the placement address if X is greater than its current value. This enables the workspaces of subroutines to share store automatically, provided they are declared in a bottom-up order. (SIR subroutines are not usually recursive). For example, if the labels LOGZ & EXPZ are placed after the last workspace of subroutines implementing LOG & EXP respectively, then the link and other workspaces used by a POWER subroutine (which calls LOG & EXP) can be located by “^LOGZ,EXPZ”. Subroutines which call no others simply all overlap their workspaces onto the start of the data area. Yes, this scheme goes against the Elliott library convention of placing the subroutine link and its entry point in adjacent locations, but in Airborne Computing Division we always prohibited mixing code and variable data in systems needing a run-time program sum-check. It is based on a patch which I wrote for the native two-Pass SIR and which I used to good effect when I was implementing BASIC in 1977.
Turing-Welchman Bombe — Roger Johnson, John Harper
The Bombe continues to run well with very few issues thanks to the work of the Engineering Team led by Mike Hillyard.
Members of our demonstration team, coordinated by John Pickles, are on site every day that the museum is open to visitors (just three days per week in January). Visitor numbers remain disappointing due to Covid. Various visits by school and corporate parties continue (mostly when the museum is closed to the public) and the team provides Bombe demonstrations for them.
Progress with the Drunken Drive Project has been slow due to problems with team members having to isolate or otherwise having problems due to the pandemic. Also, small parts have been delayed in the post, in one case, seven days.
However, we are just one part away from all parts being hardened and/or riveted. Hardening, in particular, is best left as a full batch because we are charged a fixed amount almost irrespective of the number of items.
We are hopeful that soon we will be able to make good progress with team members raring to go.
Analytical Engine — Doron Swade
Last November I reported on the successful completion of the first draft technical description, by Tim Robinson, of the Analytical Engine designs. Since then we have initiatives underway to increase the size of the team to take this work further. We have a separate initiative to view working papers donated by Anne Bromley, Allan Bromley’s widow. Allan Bromley, Australian computer scientist, died in 2002. His work was the first attempt to study the detail of the Analytical Engine designs. Our study of the new Bromley papers will enable us to assess of the scope and depth of his researches and to identify where our new studies have advanced understanding of the Analytical Engine designs.
EDSAC — Andrew Herbert
Work continues with commissioning. The main areas of focus are the arithmetic unit, main control, store addressing and delay line stores.
The interface between main control and the arithmetic unit is now stable and we are working through the arithmetic unit function-by-function. We have suffered some problems with order decoding, usually easily resolved by minor adjustments in signal levels. We have also experienced some “infant mortality” in 50 year old “new old stock” valves leading to short circuits and burned out resistors. We have also found some synchronisation problems with the microprocessor-based delay line emulators used presently for the Accumulator register.
In store addressing we have been working on making all 64 main store tanks addressable – at present we are using just one tank to run short test programs. Once addressing is fully working attention will turn to commissioning the Transfer Tank and separating the Main Input and Output Buses to allow writing back to store and short number operations. A Transfer Unit chassis was been constructed some time ago but needs updating in the light of improved understanding of the requirements and operating parameters.
The delay line team have made a start on replacing Digital Tank Emulators in the short tanks (registers) by nickel delay lines. Thus far the multiplier register is operating using a delay line and its performance is being monitored.
The teleprinter interface is under construction and test as it will soon be helpful to be able to produce diagnostic messages from more comprehensive test programs.
A simple assembler has been written in Python to allow for speedy writing of test programs. it produces binary output suitable for loading directly into the store via our Signal Sequence Injector boxes. The CCS EDSAC emulator has been updated to allow loading of such images and improved checking functions so that test programs can be verified before being used to torment EDSAC itself.
Software — David Holdsworth
Free-standing Webserver Emulation System
In November, I said:
I have plans in my head for a general mechanism that will allow me to embrace KDF9 and others. Installation of the package will only involve unzipping, and then running C and possibly Ada compilations..
I have now built this and it has been shown to work on Windows 10, Linux (ubuntu 18), and Mac OSX (with Andrew Herbert’s help). There are some decisions needed on packaging, but the installation merely involves un-ZIPping and then execution of a compilation script which is merely a sequence of gcc commands. On a Windows system it can all be done without resorting to the command window. (For reasons which remain a mystery to me, many Windows users view the command window with horror/disdain/incomprehension.) I am now applying the same techniques to other machines.
Any volunteers who would like to try it on their own systems, please get in touch.
During work on the free-standing system, the warnings generated by versions of gcc showed up a typo in my implementation of floating point, in the shape of a misplaced bracket. This turned out to be immaterial as the precedence rules of C turned out to give the intended result. However, this exercise reminded me that my floating point implementation did not round. At the time it was written, the need was to be able to validate the emerging copy-typed Usercode of the Whetstone Algol system, and floating point rounding went onto the back burner, where it stayed.
In revisiting the documentation on rounding, I completely failed to find a statement about rounding with the FLOAT instruction which converts fixed point to floating. KDF9’s 48-bit fixed point number will not necessarily fit into a floating point number with a 40-bit mantissa without rounding becoming an issue. It is for this reason that KDF9 Algol restricts integers to 39 bits + sign.
Bill Findlay’s ee9 has always done this sort of thing properly.
Data Recovery — Delwyn Holroyd
After a long hiatus, Rod Brown & I spent a day at TNMoC working on the ICT 1301 tape deck which readers may remember was moved into the Large Systems Gallery just before lockdown in 2020. After refurbished mechanical parts were replaced and some pinch roller adjustments we were able to feed tape both forwards and backwards.
The next step was to look at the signals coming off the read pre-amplifiers to check all channels were working correctly. The output from each channel is differential but it may be better to capture only a single-ended signal, subject to tests to see whether the observed noise is common mode. Despite the noise, there is still plenty of signal so I don’t anticipate a problem. Now that we have a working tape deck, the ball is in my court to finish the data capture system.
2966 — Delwyn Holroyd
Over recent months there have been a few more problems with the emulated disc system, similar to those I reported earlier last year. That said, the system frequently runs for a full eight hour shift without any problems, so it remains very difficult to debug.
A new issue has been observed several times on one of the 7181 VDUs, in that the screen fills with random characters. On one occasion this was seen to happen when a visitor knocked into the table on which it stands. All attempts to reproduce the effect have been in vain!
ICT/ICL 1900 — Delwyn Holroyd, David Wilcox
PF56 Emulator (2812/7903)
David Wilcox reports he has made considerable progress with his PF56 DDE implementation. To resolve compatibility problems between our version of ESAZ (the 1900 program for DDE testing) and SA0ADD0A (the PF56 test application), he has produced his own version of the latter, named SA0ADD0B. This is now running 17 of the 20 tests in ESAZ although a little more work is required to iron out some tight timing issues. He hopes to finish the remaining tests soon and will then start soak testing. At the moment this version of DDE will only run as a 2812 (disc controller) as he has not yet started on the code converter required for 7903 (comms controller).
SSEM — Chris Burton
Several volunteers have been cautious during the current Omicron wave, so that, while the Baby has been staffed three days a week, the safety rule that “Two Volunteers Must Be Present” means it has not always been permissible to switch on. The number of museum visitors has also been relatively low. On the occasions when the machine has run, it has been reasonably stable despite routine faults like loose connections and failed valves.
Harwell Decatron — Delwyn Holroyd
The machine continues to work well and is in frequent use both on public opening days and for education and corporate groups. Since my previous report the only maintenance required has been the replacement of a few more store anode resistors.
Meanwhile in the power supply our early warning system (consisting of two small torch lightbulbs) is showing that one pair of EL360 series regulator valves will require replacement in the near future. These valves run in parallel at almost their maximum power rating which means they already have a shorter life than others. When they start to age, the current normally shared equally between them begins to go out of balance (shown by the relative brightness of the bulbs), thus causing one valve to exceed its rating and hastening the end.
Society Email Addresses — Dik Leatherdale
If you read this edition of Resurrection right to the bitter end, you may notice that the email addresses of many of the principal committee members of the Society have changed. We are in the process of moving to role-based addresses which will not need to be changed when the personnel change. Of course, these are not real email addresses, they are forwarders and at least one of them maps onto more than one real recipient. If the initial tranche of addresses works out, we’ll extend the scheme to the various project leaders at a later date.
The latest work from the pen of Herbert Bruderer concerns pocket “calculators” from ancient Rome. Go to ccsoc.org/bru5.htm for more information.
Prof. Tony Davies writes –
The first six of the IEEE Region 8 HISTELCON conferences have their presented papers in the IEEE Xplore database. Until recently, only the titles and abstracts were open to view by all, to see the full papers required access permission to IEEE Xplore, which is extremely costly, even for IEEE members.
Now, after some lengthy negotiations the full papers have been made freely available to all persons, forever.
These are the conferences involved:
The 7th HISTELCON was held in Moscow last year (18-20 August 2021) – entirely by ZOOM. The papers are still being prepared for submission to IEEE Xplore, and initially the full papers will be available only for payment. However, negotiations to make them free to all will be attempted.
The 8th HISTELCON is planned for Florence,Italy, probably 8-9 September 2023.
The Curta Caculator
Herbert Bruderer also writes to correct an error in the last edition of Resurrection. The Curta calculator was produced in Liechtenstein, not in Switzerland.
The CDC 7600
Again, in the 50 Years Ago column of Resurrection 96 we rather oversimplified the architecture of the 7600 and its satellites.
Robert Watts writes -
The first paragraph is about the Computer Technology batch stations purchased by CERN. This was one of the two key projects I was hired to work on. CERN figured it could do better and cheaper than to buy CDC 200 User Terminal (Remote Job Entry – card reader/lineprinter/monitor/keyboard [ed]) stations. It turned out to be more difficult than anticipated!!
Also there is a small error in the Computer Weekly reporting. Those batch stations were not connected to the 7600 but to the 6600. There were no multiplexing devices directly connected to the 7600, and Intercom (interactive terminal access [ed]) only ran on the 6000 Series machines. But no doubt it made the article sound more impressive!!
Readers wishing to contact the Editor may do so by email to
Members who move house or change email address should go to
Queries about all other CCS matters should be addressed to the Secretary, Rachel Burnett at , or by post to 80 Broom Park, Teddington, TW11 9RR.
The Arcturus 18C Minicomputer
In response to our plea for information on the Arcturus 18C minicomputer, we have received two emails, the first from Iain Moffat who provides evidence for a London address in Feltham (just behind what would later become ICL’s FEL02 office) in addition to the address in Blandford Forum which we already knew. Iain also speculates that there may have been some link to the Royal Signals regiment based in Blandford Forum to this day.
But Michael Kay also writes –
Your photograph of the Arcturus 18C reminds me of the machine on which I wrote my first program. I was a sixth-former in Farnborough, Hampshire in 1968/69, and I believe we acquired the machine on loan, as a prototype from a local start-up company who thought there might be a market for computers in schools. Its only input device was a row of 16 switches, and its only output device a row of 16 lights. It had no hardware multiply instruction, so I wrote a program to implement multiplication by repetitive addition. It was possible to single-step through the program for debugging, and by doing so I managed to convince myself, and my physics teacher, that the inevitable bug was in the hardware and not in my code. The machine was sent back to be fixed, but we never saw it again. I remember being told the company developing it was local, and when we failed to get the machine back after returning it, we were told that the project had been abandoned.
In all probability this was not the machine in question but [ed] can’t resist the observation that of such stuff are future ICL Fellows made.
The Manchester University CDC 7600/ICL 1906A Installation
Peter Allan asks for some information on this unusual 1970s system in which the 1906A front-ended the 7600 acting as a job scheduler and file and peripheral server, the usual arrangement having been a CDC 7600/6400 configuration. Your editor knew software engineers from either side of the project who regarded it as “a triumph” or “a disaster”. You are invited to speculate which was which.
Jim Brookes, however was heavily involved and has much useful documentation which he has promised to provide once he has recovered from a recent house move.
Alan Turing and OxfordJonathan P. Bowen
Did Alan Turing ever visit Oxford? He is well-known for his connections with the University of Cambridge, Bletchley Park, the National Physical Laboratory, and the University of Manchester, but there is no known written archival record of him ever visiting Oxford, despite it being the location of the University of Oxford, traditionally a rival of Cambridge. However, surely he must have done so.
As a computer scientist, I first became fully aware of Alan Turing’s achievements during the mid-1980s on reading Andrew Hodges’ magnificent 1983 biography (Alan Turing: The Enigma), later reissued for the Turing centenary in 2012. Hodges is an Oxford-based mathematician who has also run a website on Turing since the early 1990s (www.turing.org.uk). I myself established the Virtual Museum of Computing online around the same time while working at the Oxford University Computing Laboratory. We collaborated and even published together on these efforts. We have since maintained contact periodically and Andrew Hodges has told me that he knows of no record of Turing having visited Oxford.
However, Hodges’ biography does include mention of some visits. In summer 1924, the Turing family stayed in Oxford, when Alan Turing would have been 12. This was probably a nostalgic visit for his father, Julius Mathison Turing (1873-1947), who was a scholar at Corpus Christi College, Oxford, and received his BA degree in 1894. One might expect Alan Turing’s father to have been keen for him to go to Oxford, but Turing’s close school friend Christopher Morcom (1911-1930) won a scholarship at Trinity College, Cambridge. Very sadly, Morcom died of tuberculosis before he could attend university, but this no doubt influenced Turing to apply for a scholarship at Trinity College too. In the event, Trinity did not award Turing a scholarship, but his second choice of college, King’s College, Cambridge, did. With hindsight, Trinity College may well regret this decision! On the other hand, without the influence of Morcom, Turing may have followed his father to Oxford. Later in 1941, during World War II, Hodges notes a weekend visit to Oxford with Joan Clarke (1917-1996), who worked at Bletchley Park and was briefly engaged to Turing, to see Clarke’s brother, Martin Clarke, previously a Fellow at King’s College.
While most of the Government Code and Cypher School (GC&CS) activities during World War II were based at Bletchley Park from 1941, the security section was located at Mansfield College, Oxford, preparing codebook and re-enciphering tables, largely printed by the nearby Oxford University Press. One of the reasons for locating operations at Bletchley Park was that it was conveniently positioned about halfway between Oxford and Cambridge on the “Varsity Line”, a railway line that ran between the two. Bletchley railway station was very close to Bletchley Park, making visits to Oxford and Cambridge easy. One could imagine the possibility of Turing visiting and helping with the operation at Mansfield College in Oxford during the war.
After the War, Turing also periodically visited David Champernowne (1912-2000), a former mathematics scholar with Turing at King’s College, who was later a fellow of Nuffield College, Oxford, Director of the Oxford Institute of Statistics from 1945 to 1948, and Professor of Statistical Economics at the University of Oxford during 1948-1959, before he returned to Cambridge. It is likely that at least one of these meetings was in Oxford. In 1948, they collaborated on an early chess-playing program, Turochamp.
In 2012, I co-organised a celebration of Turing’s centenary at the Oxford University Department for Continuing Education, in parallel with events in Cambridge, Bletchley Park, Manchester, and elsewhere. Why should Oxford not celebrate such an esteemed scientist even if there was no reliably recorded connection with Oxford? Since then I have collaborated on a major book on Turing’s work (The Turing Guide), initially based on chapters by presenters at the Oxford, Cambridge, and Bletchley Park centenary meetings. This finally appeared in early 2017, published by Oxford University Press, with three of the four main authors (the philosopher Jack Copeland, the mathematician Robin Wilson, and me as a computer scientist) being Oxford graduates.
Turing and Eisenhower in Oxford
During this time, I periodically asked during talks on Turing whether the audience knew of any evidence of Turing ever having visited Oxford, never really expecting to get a response. However, on one occasion I did. After a talk at Southwest University in Chongqing, China (where a bust of Turing can be found outside the computer science department), a former Oxford University Computing Laboratory colleague of mine, Jim Woodcock, now Professor of Software Engineering at the University of York, answered positively. He told me of a social event he attended at University College, Oxford, my own college as an undergraduate. There he met Herbert Hart (aka H. L. A. Hart, 1907-1992), who worked at Bletchley Park during World War II and was later Professor of Jurisprudence at Oxford during 1952-68, following on in the post from Arthur Goodhart (1891-1978), who was later Master of University College. He recounted a visit to University College, for a lunch during the War by General Dwight D. Eisenhower (1890-1969). There was an entourage including people working at Bletchley Park, with Hart himself and, he recalled, Alan Turing, among their number. Using British innuendo, he noted that Turing (as a gay man) made an inappropriate remark about one of Eisenhower’s aides, which seems to make the story ring true.
Subsequently I contacted the University College archivist, Robin Darwall-Smith, to see if there was any record of such a lunch at the College. Sadly, he could not find any written record in the archives, but he noted that this was not unusual during the War and that the story sounded very plausible to him. Arthur Goodhart, an Anglophile American, and Sir William Beveridge (1879-1963), Master of University College, who would probably have loved to meet Eisenhower according to Robin Darwall-Smith, would both have been at the College at the time.
Perhaps, I thought, the Eisenhower archives and diaries, available in the USA, could reveal more of his wartime movements while in England. Certainly, he stayed in the village of Sonning, on the River Thames not far from Oxford during the period before the Normandy landings (D-Day) on 6th June 1944, where I was also a resident for a while. So I contacted the archives about any known visits by Eisenhower during the War and very quickly received a helpful reply. There are records of him visiting Oxford on two occasions to see military colleagues in hospital. Valoise Armstrong, archivist at the Dwight D. Eisenhower Presidential Library and Museum in Abilene, Kansas, wrote:
I have conducted a search of the chronology of General Eisenhower’s activities during the war for any visit he made to Oxford. I found only two instances, both were visits he paid to colleagues who were ill. On October 1st 1942 he visited General Walter Bedell Smith in the hospital in Oxford. On April 16th 1944 he paid a visit to the hospital to see Col. James Gault. I did not find any record of a more formal or official visit or any mention of a trip to University College. It is possible that he had lunch somewhere in Oxford on these afternoons, but we do not have records of his meals on those occasions.
General Walter Bedell Smith (1895-1961) was a senior officer of the US Army who served as Eisenhower’s chief of staff at the Allied Forces Headquarters during the Tunisia campaign of November 1942 to May 1943 and the Allied invasion of Italy in 1943. Brigadier Sir James Frederick Gault KCMG (1902-1977) was a British military assistant to Eisenhower. For both dates when Eisenhower is known to have visited Oxford, it would have been possible for Alan Turing to be in Oxford as well. Turing visited the US during November 1942 to March 1943 (as well as studying at Princeton University for his PhD in the 1930s), so he certainly had American connections.
Of the dates when Eisenhower was in Oxford, 1944 is perhaps more likely as a rendezvous involving Turing. General Eisenhower certainly appreciated the work of Bletchley Park in the lead up to D-Day. Although he was not able to visit Bletchley Park afterwards, on 12th July 1945 he wrote a secret letter to Sir Stewart Menzies (1890-1968), the wartime chief of the British Secret Intelligence Service (aka MI6), giving thanks for the codebreaking work at Bletchley Park. Since 2016, this has been on display at the museum now located at Bletchley Park.
One of Eisenhower’s aides, Harry Butcher (1901-1985), kept a diary of his wartime experiences with Eisenhower over three years, which was later published in book form. 16th April 1944 was a Sunday and the diaries mainly cover weekdays, so unfortunately the book is silent about Eisenhower’s activities on this date. Robin Darwall-Smith noted that nobody signed into lunch in Hall at University College on this date. Could there have been a secret lunch in the Master’s Lodgings instead?
Robin Darwall-Smith also let me know that Beveridge’s archives are held by the London School of Economics (LSE), where he was Director during 1919-1937. Thus, I registered for an LSE library card to view the archives there. In particular, I noticed that the archives included Beveridge’s pocket diaries and the visitors’ book at the Master’s Lodgings. On viewing the contents of the archives, the pocket diaries were very disappointing with very few entries. However, the visitors’ book was much more interesting. For the date of 16th April 1944, there was a list of names, including locations in the USA such as New York City, Philadelphia, Wayne (Pennsylvania), Spokane (Washington), and Knightstown (Indiana). But there were no names that linked with Eisenhower (as confirmed by Valoise Armstrong, the Eisenhower archivist) or Turing and Bletchley Park. However, two names were a particular puzzle, “Hans Baumann – Man”, with a rather indistinct affiliation, and “Margaret Stewart” immediately below. After some time deciphering this, it turned out to be Picture Post, the photojournalistic magazine that ran from 1938 to 1957. Hans Felix Sigismund Baumann (1893-1985), aka Felix H. Man to disguise his German heritage, was a pioneer photojournalist.
So, I thought, was there an article in Picture Post after this event? Picture Post is accessible online but at a fee. Fortunately, I have a long-time friend – Andrew Robinson, editor of the book The Scientists, on 43 leading scientists of all time, including Alan Turing – who subscribes to the London Library, which provides access to copies of Picture Post. He kindly found the article on this event, including names of people involved matching the signatures in the visitors’ book, with text by Margaret Stewart and photographs by Felix Man. It covered Sunday tea with Sir William and Lady Beveridge for American officers billeted in Oxford before D-Day, noting that they held such tea parties on Sundays when at the Master’s Lodgings in Oxford.
The tea party on Sunday 16th April 1944 at the University College Master’s Lodgings and reported in Picture Post did not involve Eisenhower or Turing. However, it demonstrates that Beveridge was in Oxford on that date. We also know that Eisenhower was in Oxford on that date. We have no record of their lunch arrangements. We have a verbal report that Eisenhower had a lunch at University College with members of Bletchley Park, including Alan Turing. There is no record of anyone signing in for lunch in Hall at University College on 16th April 1944, so a secret lunch at the Master’s Lodgings is entirely possible. Thus, there is significant circumstantial evidence that Turing may have visited Oxford with Bletchley Park colleagues and a lunch at University College with Eisenhower arranged for them, since it is known that Eisenhower appreciated the work of Bletchley Park very much. But it would be nice to have some written evidence for this, a silver bullet of proof, perhaps in a letter somewhere recalling the event.
Turing connections in Oxford
Whether or not Turing visited Oxford in any significant way during his lifetime, his legacy certainly lives on at the University. Turing’s PhD student at Cambridge and friend Robin Gandy (1919-1995) joined the Mathematical Institute at Oxford in 1969 as Reader in Mathematical Logic. He was also a Fellow at Wolfson College, Oxford, and generously left his estate to the College, where there is now a residential block, the Robin Gandy Building, named after him. The College celebrated his centenary slightly late on 22nd February 2020 with a one-day colloquium. Dana Scott studied for his PhD under the Princeton mathematician Alonzo Church (1903-1995), the same supervisor as Turing. Scott’s research straddled mathematics and computer science at the Mathematical Institute and he received the ACM Turing Award, the highest international award for computer science, in 1996. One of Scott’s PhD students at Oxford, Jack Copeland, is now a leading Turing scholar. Roger Penrose OM FRS, Emeritus Rouse Ball Professor of Mathematics at Oxford, has been influenced by Turing’s ideas on artificial intelligence (specifically the Turing test) and computability (specifically the Turing machine) in his book The Emperor’s New Mind. His PhD student and subsequent research collaborator at Oxford, Andrew Hodges, is also Turing’s definitive biographer. More recently, Philip Maini FRS, at Oxford’s Mathematical Institute, is a leading researcher in mathematical biology, a field that Turing arguably founded with his 1952 paper on morphogenesis and Turing patterns, published not long before he died.
Oxford has also been a leading centre for formal methods, originally at the Programming Research Group. This was founded in 1965 under the leadership of Christopher Strachey (1916-1975), a colleague of Turing at Manchester, and then led by Tony Hoare FRS after Strachey’s untimely death, applying mathematical approaches to reasoning about software (and hardware). This is a field that Turing pre-empted, in particular with a short and lesser-known foundational paper in 1949 that has subsequently been examined in detail. Samson Abramsky FRS FRSE now holds the Christopher Strachey Professorship at the Oxford University Department of Computer Science. He has made contributions to the foundations of computer science and has been influenced by Turing’s legacy. In recent years, data science researchers at Oxford have been highly involved with the Alan Turing Institute in London. So certainly overall, Turing’s spirit and influence continues at Oxford in several fields.
Alan Turing has blue plaques celebrating locations with which he is associated, such as his place of birth in Maida Vale, west London, and his place of death in Wilmslow, Cheshire, at his home south of Manchester. There is no plaque for Turing in Oxford, for obvious reasons as shown in this article.
However, his wartime fiancée and colleague Joan Clarke (later Joan Murray through marriage) moved to Oxford during her retirement and lived at 7 Lakesfield, Headington Quarry, on the outskirts of Oxford, from 1991 until her death in 1996. On 27th July 2019, the Oxfordshire Blue Plaques Board unveiled a plaque to record this connection. So, there is at least now a transitive plaque for Alan Turing’s presence in Oxford.
A Turing seminar in Oxford
In 2020 during the Covid pandemic, I gave an online talk based on my research to that point on evidence of Turing having visited Oxford to the Oxford History of Mathematics Forum, which normally meets at The Queen’s College, Oxford. Christopher Hollings, a leader of the Forum and a College Lecture at the College, noted a short piece he had read in The Queen’s College Record by the topologist Professor Ioan James FRS, who matriculated at Oxford in 1946, stating: Another Mathematics lecturer worth mentioning was Alan Turing, who told us about what became known as the Turing machine.
James spent some time as a student in Cambridge due to illness as well as Oxford, so it was not completely clear whether his memory was on a lecture in Oxford or Cambridge. James was later an academic at St John’s College, Oxford, but by now is retired of course. Christopher Hollings suggested writing to James, which I duly did with a draft of this article. I thought I would hand-post the letter since I live not far from St John’s in Oxford. However, on arriving at the College, it was completely barred to visitors during the pandemic, even to post a letter. So, I put a stamp on my envelope and entrusted it to the Post Office. I heard nothing for several months, but in February 2021, a hand-written letter arrived at my home. In it, James describes what he calls the “old guard”, including Theodore Chaundy (1889-1966) of Christ Church, Oxford, and a “new guard”, led by the statistician and mathematician David Kendall (1918-2007), from 1946 a tutor at Magdalen College, Oxford, and later a professor in Cambridge from 1962. Kendall was a contemporary statistician with David Champernowne, another colleague of Turing as mentioned previously, and the two overlapped at Oxford in the late 1940s and 1950s.
Apparently, Turing lectured at one of Kendall’s seminars in Oxford, with a small audience, including James. So, there is at least one living memory of Turing having visited Oxford University in an academic capacity.
Finally, this is a plea to any reader who may have further information or know someone with additional knowledge on any further Turing-Oxford connections to get in touch with the author by email at . It would be especially wonderful to confirm the Eisenhower – Turing visit if there is any further evidence to be found. But at least there are indications, if only verbal and circumstantial, that Alan Turing did visit Oxford on a number of occasions, and written evidence for at least one occasion.
Thank you to the following for help with this article: Valoise Armstrong, Jack Copeland, Robin Darwall-Smith, Tula Giannini, Andrew Hodges, Christopher Hollings, Ioan James, Tomoko L. Kitagawa, Andrew Robinson, Jim Woodcock, and the Porter at Mansfield College, Oxford. All photographs are by the author.
This article has been edited to accord with Resurrection’s house style. The original version which is fully referenced can be found at ccsoc.org/turingox.pdf [ed].
A Life of Calculation (Part 2)Ron Weeden
The HP 9815 programmable calculator was also introduced in 1975, designed and manufactured in Fort Collins, Colorado. Steve Wozniak was on the HP 9815 design team. Before leaving HP to set up Apple with Steve Jobs, Wozniak showed his HP bosses his computer on a board, the prototype of the Apple, but HP dismissed it as rubbishy engineering. HP was not then interested in the mass consumer marketplace but only in ‘Rolls Royce’-type customers wanting quality engineering and willing to pay the price. Such customers could not afford the millions of pounds that could be lost through having a computer go down and needed very long term reliability of such machines. Sometimes such a machine could be controlling the manufacturing process of an entire factory or of a major department of a factory or driving numerically-controlled machine tools operating on extremely expensive pieces of metal. Six HP 9825s were used to control the machining of Uranium 235 hemispheres at the Royal Ordnance Factory, Burghfield near Aldermaston in Berkshire via enormously expensive numerically-controlled machines.
Hewlett Packard desktop computers used gold-plated circuitry and were oven baked, drop tested, vibration tested and freeze tested many times.
The operating system of all the above computers was held in ROM and the processors used binary coded decimal (BCD) arithmetic, each decimal digit being encoded in four bits for the decimal digits 0 to 9. Typically results were displayed to 12 decimal digits although the internal calculations were usually performed to about 16 decimal digits and then rounded off.
The HP 9845A was introduced in 1977 using the same HP 16-bit BCD microprocessor as in the HP 9825. It was soon followed by the more powerful HP 9845B using the same processor. The HP 9845 incorporated the first incarnation of HP’s extremely powerful Rocky Mountain BASIC which totally eclipsed the primitive forms of BASIC which had been around for many years. It was a full blown properly high-level language incorporating all the features required for professional structured programming but with an ease of use such that any engineer could happily program it. I should add that a similar form of BASIC had previously been used for writing and running local programs in HP 2648 terminals; these also had both text and graphics facilities on their inbuilt CRT screens and could be used to programmatically control Hewlett Packard plotters; it was called AGL (A Graphics Language) and was the direct ancestor of Hewlett Packard Rocky Mountain BASIC.
Following HP’s tradition, the 9845B had all the features required for easy instrumentation control and monitoring and of course it had very good inbuilt graphics capabilities as well as supporting all of HP’s own range of computer peripherals and many others. It also boasted an inbuilt thermal line printer running at 480 lines per minute; that was twice the rate of the HP 9866 introduced in Nov 1972 for use with the HP 9830 and which could also be used with the HP 9825 and 9831.
The CRT was usually monochrome but, at greater cost, it could be a colour one. The Rocky Mountain BASIC was supplied in ROM, with optional extensions in ROM which the user could insert into slots in ROM drawers on the left and right hand sides of the desk top computer. The HP 9830, 9820, 9821 and 9810 had also used plug in ROMs to extend the capabilities of their intrinsic ROM based systems.
The first HP machine not using HP’s own microprocessors was the HP Series 200 machine known as the 9826 introduced in May 1981. It used the Motorola 68000 8MHz processor and had an inbuilt CRT and an inbuilt 5¼” floppy disc drive. It was followed in 1982 by the HP 9836 which had a much larger (non inbuilt) CRT and two 5¼” floppy disc drives. In 1983, the model 9836C was introduced which boasted a colour CRT but otherwise looked the same as the monochromatic 9836. The CRTs had both text and graphics capability. In the case of the model 9836, the text capability was 25 rows of 80 characters but, in the case of the model 9826, rather smaller row and column-wise.
Meanwhile, in November 1982, HP introduced the extremely popular model 9816 at the same time as introducing to the world the first 3½” floppy disc drives. HP had given a contract to Sony for 10,000 such units. The 9816 was extremely compact; the memory, the processor, etc, all being housed within a small 25 line by 80 character wide monochromatic CRT. The twin 3½” floppy disc drives were designed to be of exactly the same width and breadth as the 9816 which sat on top of the 3½” drive box. A relatively great number of HP 9816s were sold. It was to be many years before 3½” floppy drives commonly replaced the 5¼” ones on the majority of PCs from other manufacturers, despite the enormous disparity in reliability between 5¼” floppy discs and 3½” ones. Like those associated with all the HP series 200 systems, the CRT of the 9816 had both text and graphics (400*300 pixels) capability and 80 column by 25 rows of totally scrollable text.
I might add that exactly the same screen was used to house an Intel 8-bit microprocessor instead and together with the 3½” dual floppy drive was sold as the HP 120 PC running the CP/M operating system. It took the place of HP’s earlier PCs which were considerably greater in volume and desk space requirements. In the case of both the HP 9816 and the HP 120, a small keyboard was externally attached. All of this was in the days before mice were introduced and before Windows environments existed. The first Apple Windows system was introduced on the Apple LISA in 1983. Incidentally, the first three Apple LISA machines to reach the U.K. were delivered respectively to IBM Research Labs. at Winchester, Hewlett Packard software development centre at Crowthorne and to Digital Equipment Corporation in Reading.
For the HP Series 200, there was also a 9888A expander box which could accommodate up to 16 interface cards of the same kind as fitted into the back of the series 200 and 300 systems. We called those cards DIO1. There were wider cards for mother boards and for graphics which we called DIO2.
Later models of the Series 200 were the 9836, the 9920, 9817 and 9837. From the beginning of 1983, the 9826, 9836, 9920, 9816, 9817 and 9837 became known as the models 226, 236, 220, 216, 217 and 237. They all used Motorola 680XX 16-bit microprocessors. All the HP Series 200 systems could run a choice of HP Rocky Mountain BASIC or HPL or HP Pascal as self-contained integrated program language/operating systems. In the 216, 226, 236 there was a choice of having the Rocky Mountain BASIC or HPL in ROM, instead of being supplied on 3½” or 5¼” diskettes. This meant that boot-up was virtually instantaneous because there was nothing required to load into memory before the system was ready for use, other than any program which one wished to run. The HPL and HP BASIC language operating systems were both written in a language called Modcal which was a modified form of Pascal containing language enhancements which made it suitable for writing the code for operating systems, unlike the standard form of Pascal which was thought unsuitable for that kind of thing.
The first of HPs Series 300 models were introduced in July 1984, those being the Model 310 using Motorola 68010 and Model 320 using Motorola 68020. They were soon followed by other models using more and more powerful Motorola microprocessors. Thus there were models 318, 319, 330, 332, 340, 345, 350, 360, 370, 375, 380 and finally the 362 and 382 using Motorola 68040 processors. The Series 300 machines took the place of the Series 200. The series 300 machines continued in production until about 1993, the final model being the R382; a rack mounted form of the 382 which also incorporated a touch screen.
Going back to November 1982, HP also introduced its Dawn machine, at first known as the HP 9000 , but after introduction of further models becoming the HP 9020. The following year it became known as the HP 520 as part of the newly designated HP Series 500. The HP 9000 Dawn machine was the first commercially-marketed computer to be based upon a 32-bit microprocessor. Indeed it could have three such processors all operational at once. The 32-bit processor was known as the FOCUS chip and was designed and produced in HP’s own microprocessor plant in Fort Collins, Colorado. All series 500 machines used that same microprocessor. Specimens of the Dawn machine (9000 = HP 520) are at the National Museum of Computing and also at the Computer Museum at Swindon, where there is also an extensive collection of the HP series 200 and 300 systems with numerous peripheral devices such as printers, plotters, digitising tablets, disc drives, etc, etc).
A transparent Perspex block containing one of the HP Focus 32-bit processors is held in the Museum of Computing at Swindon. The details of its provenance are engraved upon its base. Such a sample was issued to every HP employee worldwide in November, 1982.
The HP Dawn machine was designed to run an enormously enhanced form of HP Rocky Mountain BASIC which included an inbuilt database and inbuilt 3D graphics in the language/operating system. The Rocky Mountain BASIC actually ran on top of what we called the SUN operating system which was fully multitasking. We also allowed a choice of using Hewlett Packard’s form of Unix, Hpux instead once again running on top of the SUN OS. This SUN OS had nothing to do with the American workstation manufacturers called SUN. The Dawn machine had a colour screen, inbuilt 5¼” floppy disc drive and an inbuilt Winchester hard drive. Although designed as a desktop machine, it usually took two people to lift it unless the CRT was first removed. At the time of its introduction in Nov 1982 it sold for £40000. The language and operating system were normally held on the internal Winchester drive. Even the Rocky Mountain BASIC was fully multitasking with facilities for task scheduling.
The HP Rocky Mountain BASIC on the Dawn machine also incorporated a revolutionary incremental compiler. I am not aware of that technology having later been used elsewhere. The same very fast thermal printer as used in the HP9845 was an intrinsic part of the Dawn machine.
In 1979, the calculator division of HP (that producing the HP 65 and its successors in Portland, Oregon) introduced the HP 85 desktop computer which incorporated a form of BASIC in ROM but nothing like as good or as powerful as HP’s Rocky Mountain BASIC. This sold to small businesses and some scientific applications. It used a BCD processor designed and produced in HP’s calculator IC plant in Oregon. It had a small inbuilt CRT and a single mini cartridge drive. It could also have external 5¼” floppy disc drives attached. After the very popular HP 85 came the models HP 87 and HP 86, by which time the 3½” floppy disc drives were available from HP made by Sony. I should add that by comparison with the computer products of other HP divisions, the Series 80 machines (HP 85, 86 and 87) were extremely slow but they were used in applications where that was not a problem.
All HP computers were equipped with HPIB ports and could thus be attached via HPIB to HP’s extensive range of computer peripherals produced by various manufacturing divisions of HP. At one time there were four different HP divisions competing with each other producing differing ranges of printers. Other divisions designed and produced plotters and digitising tablets, whilst others produced hard and floppy disc drives. Not to mention divisions producing computer terminals for use with HP’s large 3000 Series commercial mainframe computers. Of course, the HPIB ports were also used very extensively to control and monitor electronic instrumentation; very often that manufactured by Hewlett Packard’s instrumentation manufacturing divisions, but also by numerous other electronic instrumentation manufacturers.
Hewlett Packard was set up in 1939 by Bill Hewlett and Dave Packard on the basis of obtaining a large contract for the electronic audio devices required for production of the Walt Disney film Fantasia.
For many years HP was an instrumentation manufacturing company with a great many manufacturing divisions; but about 1964 it got into computers with a small machine similar to a DEC PDP 8 using paper tape for data and program input Soon other divisions were set up to design and manufacture various other kinds of computer and eventually that grew so much that about 1989 the computer side overtook the traditional instrumentation and medical equipment sides in revenue. About 2001, the company split such that the computer side retained the name Hewlett Packard and the instrumentation side became a totally independent company known as Agilent Technology. Over the years beforehand, HP had absorbed Apollo computers and later bought up Compaq who were producers of a range of PCs. Compaq had previously been acquired by the Digital Equipment Corporation often fondly known as DEC.
HP holds many computer technology patents; even the use of soft keys is a Hewlett Packard patented technology probably since about 1972.
The first pocket calculator was the HP model 35 scientific calculator (Nov 1971) designed on the express instruction of Bill Hewlett himself by his personal design team. It included almost all the features of the original HP 9100 desk top programmable calculator, except that it lacked programmability. Next was Clive Sinclair’s cheap four-function calculator (+,-,*,/ only), several months later. The HP 35 used digit by digit methods for all its scientific functions such as LOG, EXP, SQRT, SIN, COS, TAN, ASN, ACS, ATN, etc as did the first HP desktop calculator, the 9100a, which was the source of the inspiration for the HP 35 because Bill Hewlett wanted a version of the 9100 calculator which he could put into his shirt pocket. Bill Hewlett was very much a practical engineer and would often go down to the workshops to use lathes and milling machines by himself. The HP35 was later followed by the programmable HP67 and then by a whole range of HP and Tektronix pocket calculators. The HP 67 used very tiny magnetic cards for storing its programs. The card reader was inbuilt into the body of the pocket calculator and thus took up no extra space.
HP entered the field of personal computers in about 1981 or just before with machines using the CP/M operating system and Intel processors. This was a natural consequence of HP’s production of computer terminals for customers of its large 3000 Series business computers. The same terminal screens were simply made to house an additional independent Intel microprocessor so that they could function locally without reliance on connection to an HP mainframe computer for the running of programs. These PCs were called HP125s.
IBM entered the field in 1983 with the IBM PC and created the near-monopolies enjoyed by Microsoft and Intel in the PC world of today, even though both the OS and the processor were about the worst choices that could have been made.
The advantage was not quality but cheapness because of all the home and games computers that were around based on Intel and Zilog Z80 microprocessors (Intel and Zilog having similar architectures to each other) and almost all using the very primitive forms of BASIC implemented by Microsoft which was so primitive because it was designed to use as little memory as possible; computer memory being expensive in those days. It was little wonder that the name BASIC as a programming language raised instant hackles among computer professionals; although it was only Microsoft that made it as awful as it was; the original Dartmouth BASIC invented in 1964 at Dartmouth College for use via computer terminals connected to large mainframe computers was very good and avoided the use of enormous bundles of Hollerith punched cards which was otherwise the usual way of submitting programs to a mainframe computer of the time.
The inventors of Dartmouth BASIC, Kemeny and Kurtz, eventually produced a book and a diskette called TRUE BASIC to show what it should have been like and indeed had been like as designed and used at Dartmouth College in the USA and how debased and diversified into many incompatible variants the language had become in the hands of Microsoft. I ought to say that before the time of Microsoft, there were many computer time share companies selling time on mainframe computers via computer terminals and those all used various forms of BASIC as the programming language.
Microsoft BASIC came in as many variations (dialects) as there were makes of home computer to use it. Most of the burgeoning of home computers for games use came after 1975; probably the most common was the Sinclair ZX81, although the Motorola based Atari range was also very popular. There were so many varieties of home computers for games that I could never keep up with remembering them all. However, one significantly different one used a Motorola processor together with the Microware OS9 operating system with BASIC09 which was a very sophisticated and powerful form of BASIC far better than those produced by Microsoft or indeed by anyone else other than HP.
Today’s PCs would be amazingly better had it not been for the choices made by IBM back in 1983. Microsoft Windows still suffers internally from the consequences of those choices. Had IBM chosen to use 68XXX Motorola microprocessors there would have been an intrinsic capability for growing PCs to full 32-bit capability because that was intrinsic in the architecture of a Motorola 68XXX microprocessor. A much better operating system than MSDOS would then have been used which would have had amazing consequences for the subsequent history of computing. As it was, PCs remained unnecessarily very slow, primitive and unreliable for far too long. It is a very great shame that Microsoft and IBM were ever involved.
After the advent of pocket calculators, the same burgeoning of varieties happened as with home and games computers. It seemed that almost every man and his dog were producing pocket calculators in their back garden sheds using cheap Japanese and American components.
Hopefully, this will give a total overview of the nature of personal calculating and computing devices before the advent of today’s range of ubiquitous personal computers using Intel processors and Microsoft operating systems; although other operating systems are available which are all forms of Unix; e.g. Xenix, Redhat Linux and Ubuntu Linux, the last of which is open source. A form of HP’s Rocky Mountain BASIC ran on Redhat Linux platforms for many years and another, almost totally compatible form called HT BASIC still runs under Microsoft Windows (at least up to Windows 7). That version is available from Transera Corporation in the USA. It was made nearly fully compatible with Rocky Mountain BASIC by a joint effort between HP and Transera software engineers in the mid 1990s. There is a UK agent, Tony Bolding, for Transera Corporation.
Incidentally, the original version of the Microsoft Windows operating system for PCs was designed by an HP team of 16 software engineers at Nine Mile Ride near Crowthorne, in Berkshire. HP and Microsoft were later sued by Apple in the New York courts for some vast amount of money but it was settled out of court for some probably very large sum.
HP started the design in May 1983 and it took our team of about 16 to 20 software engineers some 10 years to complete the project for Microsoft. Those software engineers were my colleagues although I did not become personally involved except at the very start, on the day we commenced.
HP never publicised that fact and there are extremely few people in the world who know it. It can be confirmed by looking at the financial news of the time when HP and Microsoft were sued in the New York courts; probably about 1993 so far as I can recollect.
As a matter of record I have a legal copy of the source code of Rocky Mountain BASIC and also that of Microware’s BASIC 09 with full rights to do whatever I wish with them. My hope is that they may yet be put onto an open-source platform such as Ubuntu Linux. They are certainly both worthy of preserving for posterity; they are wonderful examples of humanly-oriented software engineering. If anyone can contribute their expertise to that endeavour, I would dearly love to hear from them. Obviously, deep knowledge of the internals of the open-source operating system would be required and also experience in compiler design.
I would dearly love to see them available for use on the Raspberry Pi and similar processors. I can give full guidance on what needs to be added to those languages to bring them into the modern age.
N.B. This article has expressed the author's strongly-held opinions on various matters of historic software and hardware. Readers should be aware that neither [ed] nor the Society endorse any such views, neither does it oppose them. We do, however, believe that the author is entitled to express them as he thinks fit.
CCS Website Information
The Society has its own website, which is located at www.computerconservationsociety.org. It contains news items, details of forthcoming events, and also electronic copies of all past issues of Resurrection, in both HTML and PDF formats, which can be downloaded for printing. At www.computerconservationsociety.org/software/software-index.htm, can be found emulators for historic machines together with associated software and related documents all of which may be downloaded.
Terry Froggatt’s Resurrection 96 report on a very-long standing software error suggests that there may be other bugs from long ago, never corrected but lurking for years without being detected. I found one such in the Brooker-Morris Compiler Compiler (CC).
In Resurrection 92 I described the general structure of CC programs. For our purposes today we’ll be concentrating on the imperative, procedural language in which one might write such statements as
B2 = 8 B3 = 4 B2 = B2 + B3 | B1,B2 etc. are members of Atlas’ large set of 24- | bit integer index registers.
But first we need to understand the difference between the way that Atlas represented integer values and the way that CC did. Atlas assumed that the 24 bits of a halfword are 21 bits of a whole number and a 3-bit fraction (21+3 notation). So, adding 1 to an address gave the location of the next 48-bit floating point number or instruction. CC, however, had little use for floating point numbers and so decided on a 22+2 notation. Adding 1 to take us to the next halfword integer was more useful. So the instructions compiled for the sequence above would be
0121 2 0 4 | B2 := 4 0121 3 0 2 | B3 := 2 0124 2 3 0 | B2 := B2 + B3
which gives us the 21+3 mode answer of 6 which would be interpreted by CC as 12 – the answer we expected.
Now just to make it more interesting we might instead write
B2 = 8 B3 = 4 B2 = B2 X B3
And whereas the add instruction doesn’t care where the boundary between the whole number and the fraction is located, the multiply instruction does. So, the CC compiler doubles both B2 and a copy of B3 to bring them into line with Atlas conventions. Then it embarks on a 21+3 mode multiply and finally a one-bit right shift which gives us the right answer.
0121 2 0 4 | B2 := 4 0121 3 0 2 | B3 := 2 0124 2 2 0 | DOUBLE B2 0121 83 3 0 | B83 := B3 0124 83 83 2 | DOUBLE B83 1302 2 83 0 | B2 := B2 * B83 0163 2 0 0 | RIGHT SHIFT B2 BY 1 BIT
Or does it?
Let’s imagine what would happen if one or other (but not both) of the operands is negative. We end up with a (CC) value of 524280. How can this be? It turns out that the problem is that final right shift. The 0163 instruction is a circular shift so the least significant bit (almost inevitably a zero) ends up in the sign bit position! By the way, if we tried B2 = B2 / B3 we would generate similar code, and observe the same error.
Now admittedly my copy of CC dates from December 1962, the point at which CC started to give useful results. So perhaps a few embedded coding errors might be expected. But I borrowed a 1967 listing and lo, it’s still there! By 1967, CC would have been regarded as pretty stable, so we are on reasonably firm ground if we assume that this error was never found during the lifetime of Atlas.
The obvious solution is to replace our circular shift instruction with an arithmetic shift –
1340 2 0 1 | ARITHMETIC RIGHT SHIFT B2 BY 1 BIT
And this works well. It delivers the correct result albeit taking fractionally longer.
But a better solution is to omit the doubling of B2 and the shift. This also gives us the correct result in all circumstances and is not only quicker but more compact.
So it occurred to me to wonder how this problem has lain undiscovered for 60 years. CC was created for a very specific purpose, the implementation of compilers. So it’s likely that, in that limited sphere of activity, multiplications were quite rare and multiplications involving negative numbers, vanishingly so. It’s entirely possible that nobody ever had to multiply a negative number let alone divide one.
So have I discovered a bug even older than Terry’s? Probably. Is it the oldest bug yet discovered? Sadly, the answer is no. In 2015 Bernard Sufrin gave a splendid presentation at the Oxford conference celebrating Ada Lovelace’s 200th birthday. He asserted that Lovelace’s famous “first-ever” published program of 1843 contained a subtle but significant error.
Collapse of stout party.
Dik Leatherdale is the editor of this very publication. That’s why this sort of nonsense is allowed in these hallowed pages.
50 Years ago .... From the Pages of Computer WeeklyBrian Aldous
Cassette and New Tape from EMI: Two new high density tape products have been announced by Emitape Inc. The Datasette digital tape cassette carries 300 feet of tape and stores data at 1,600 bpi, while the type 6400 reel tape has a storage capacity of 6,400 flux changes per inch. In designing the Datasette great attention was paid to ensuring purity of recording. The casing is manufactured from high impact plastic and metallised to eliminate build-up of static charge, and all metal parts, including screws, are non-magnetic to avoid spurious recording signals on the tape. (CW 2/3/72 p8)
£1.3m RAF order for Cossor terminals: One of the largest single orders ever placed in the UK for terminal equipment has been won by Cossor Electronics. The order, which is valued at £1.3 million, covers the supply of over 1,000 terminal units comprising 450 VDUs and 572 character printers for use in a stores management system being developed for the RAF. The system is being developed at RAF Hendon and the main contract for terminal equipment has been awarded to Cossor following the success of a £35,000 evaluation order for 16 displays. As well as the visual units, which are to be DIDS 402Es, and character printers, Cossor will supply 97 fanout control units and 47 printer sharing devices. When the project is completed every RAF establishment in the UK handling stores will be equipped with a terminal. As revealed more than a year ago the system is to be driven by two ICL System 4/72 computers back-to-back, and these will provide round the clock control of stores. (CW 9/3/72 p1)
OCR font gets US acceptance: After revision by the European Computer Manufacturers Association, the OCR-B font which is widely used in Europe but has been spurned in the US, has now gained acceptance from the American National Standards Institute, an outcome which once seemed unlikely. The revised font - the second edition of Standard ECMA-11 on OCR-B - includes changes to the numeral zero and the upper-case letters O and D to make them more easily distinguishable. An important point in view of the wide use of numeric subsets is that both the old and new designs for zero can be distinguished from the other numerals at little extra cost and without significant deterioration of machine performance. The Size II font has been deleted due to a lack of interest in it, and a Size IV, which is particularly suited to credit and identity card applications, has been introduced. It has the same aspect ratio as Size I and can be derived from it by linear magnification. The original Standard ECMA-II has been the subject of an International Standards Organisation recommendation since 1969, but its use has largely been confined to Europe. In the US it was felt that the stylised OCR-A fount was acceptable and that OCR-B, which has a more conventional appearance, would not be so efficiently recognised by machines. (CW 23/3/72 p8)
Supermarket check-out terminal for Britain: The supermarket check-out is not the place to go to without any cash, or so one would have thought. But now a retail point-of-sale terminal called ASTROS has been brought to this country by MSI Data Corporation of California, at which the check-out clerk does not enter prices, but only the code for the goods. The terminal, linked to a computer, records the right price according to the number of each item purchased or its weight. Moreover, optional extras with ASTROS include an on-line weighing machine and a cheque writing unit. With the latter the customer hands over a cheque and the unit fills in the payee details, date and total and then all the customer has to do is to sign it. However, just in case anyone should be so old-fashioned as to want to pay in cash the terminal does have an electronically operated money drawer. (CW 9/3/72 p8)
BEA first customer for SITA network: With the inauguration of a data link between Rome and London, British European Airways has become the first of the world’s airlines to use the SITA High Level Network for transmitting seat reservation data. A significant aspect to this development is the fact that BEA had concluded that it was not economical to rent a dedicated line at a cost of £18,000 a year, to link the limited number of terminals at their Rome sales office to the Beacon computer centre in London. However, availability of the SITA service which provides a high-level link at a line cost of only £4,500 a year, made it economically practicable to do this. In fact, the Rome office is the first of 22 BEA sales offices throughout Europe which will be linked, via the SITA High Level Network, to the airline’s £8 million Univac 494 computer centre during 1972. (CW 9/3/72 p9)
Revolutionising the Newspapers: A cathode ray tube phototypesetting system, which, according to its manufacturers, may revolutionise newspaper production and cut composing costs, has been developed in the US by Seaco Computer Display Inc, of Garland, Texas. The Seaco system, known as the Model 1601 CRT, can be used to define the type size and heading for each story and its location. It can also place heads with the stories and ensure that they fit within the parameters defined. Advertisements are not only fully coded for typesetter output but their location is coded with exact XY coordinates. The Seaco system is capable of typesetting an entire newspaper in one single pass on the CRT in about 50% reduction. This page then has to be enlarged before printing. Cost of the new system is less than $200,000. ( CW 9/3/72 p16)
Fatigue Testing Concorde past endurance: The structures testing hangar at the Royal Aircraft Establishment at Farnborough is given over completely to the fatigue testing of a full-size Concorde assembly, but the building and the machinery it contains are built on such a vast scale that the mighty aircraft inside is lost by comparison. In order to simulate the exacting conditions of supersonic flight, the whole aircraft structure is covered with hydraulic jacks and with heating and cooling ducts to such an extent that only the wingtips and the top of the fin are visible. The hydraulic jacks apply loading stresses and the various heating and cooling ducts serve to set up thermal stresses similar to those encountered in flight. In order that structural defects shall show up first in the test rig rather than in aircraft in flight, the simulated stresses are applied three times more often than they are likely to be experienced in operation. As well as the fatigue tests being carried out at Farnborough and the in-flight tests using the prototype aircraft, static tests are being conducted at Toulouse in France where a Concorde structure is subjected to conditions 50 per cent more severe than those encountered in the air. Besides the problems of mechanical stress, the Concorde designers had to bear in mind the considerable thermal stresses set up early in any one flight because the outer shell of the aircraft heats up more quickly than the interior of the hull. Conversely, when the aircraft comes to a halt, the interior is at a higher temperature than the exterior. The shell of the Concorde is therefore continually exposed to cycles of thermally produced stress. (CW 16/3/72 p1)
Varian 620/L Aids Research into Diagnostic Pictures: Dramatic improvements in the quality of diagnostic pictures may be expected from a three-year research project just beginning in the Clinical Physics and Bio-Engineering Department of Scotland’s Western Regional Hospital Board in Glasgow. Using a Varian 620/L minicomputer, the first to come out of Varian Associates’ new Scottish factory at Donibristle, the research will aim to “clean up” pictures obtained from ultrasonic equipment, isotope scanning equipment and infra-red thermography, all techniques which have been added to the best known diagnostic aid of this kind, X-rays. Some of the early work in ultrasonics has been useful in obstetrics, as a more efficient way of diagnosing multiple births than by using X-rays, but ultrasonic diagnosis has been hampered by the poor quality of the pictures, which have a fuzziness caused by the equipment rather than by the patient. (CW 16/3/72 p20)
Powerful PDP-16M introduced by DEC: A low cost minicomputer, aimed at both OEM and end user applications in data logging, machine control and communications processing, is the latest item to come out of the Digital Equipment stable. Known as the PDP-16M, it sells for a basic price of £935. The PDP-16M eliminates the hard-wired back panel architecture of its predecessor the PDP-16 and replaces it with more flexible programmable read only memory (PROM). Commenting on the introduction of the new machine, Mr Ian Follows, PDP-16 manager for the UK, told Computer Weekly, “The 16M is four times more powerful than the PDP-8e and it is particularly suitable for dedicated applications.” (CW 13/4/72 p3)
Memorex enters Mainframe Business: A new contender has come into the mainframe business with the announcement in the US that Memorex, formerly known primarily for its disc packs and plug-compatible peripherals, is introducing two commercial data processing systems. The computers, known as the MRX/40 and 50, are the first of an IBM compatible range which will eventually comprise four machines. Both will be offered in multiprocessor configurations with emphasis on the use of large disc storage capacity and comprehensive telecommunication facilities, but with the needs of the smaller commercial user in mind. The machines are fitted with MOS semiconductor memory, the MRX/40 having a storage capacity of 16K to 64K bytes while the MRX/50 is available with 16K to 128K bytes. A full range of peripheral equipment is available and systems can be equipped with from 29 million to 232 million bytes of on-line exchangeable disc storage. (CW 30/3/72 p1)
M-way signals system now operational: The first section of the new centrally controlled and computer monitored national motorway signalling system became operational last week on 85 miles of the M6 and M62 in Cheshire, Lancashire and Yorkshire. When this national system is fully operational, 30 police control offices will be able to handle the signals on the whole of the motorway network through six unattended computer centres. The total cost of the section now open was £600,000 and the main contractors were GEC-Elliott Traffic Automation, which supplied control offices and the computer system and Plessey, which supplied the equipment on the motorways. For this section a computer centre at Westhoughton, Lancashire, containing two MARCH 9050 computers, one of them in a stand-by role, is controlling 164 motorway signals. (CW 6/4/72 p1)
IBM announces two systems to fight pollution: With pollution a burning issue these days, IBM is losing no opportunity to win any public sympathy that’s going. Following the announcement of an atmospheric model system for New York City comes news of two more systems for the ecologist. In Atlanta, Georgia, a water pollution control system based on an IBM System 7 was demonstrated at the Society of Manufacturing Engineers Convention in February. The system employs sensors which are placed in a river both above and below the plant or city. When the sensors record the approach of predetermined pollution limits, the valves on pipes leading into the river from aerating tanks containing partially processed waste are automatically closed. IBM claims that the time element is the big advantage of the system over present methods. “At present, action is not usually taken until a fish kill is spotted, or a violation has occurred. By then the harm is done”. Another feature of the system is that “the same method . . . can also be used to monitor and control air and noise pollution.” In Missouri, a research project sponsored by the US Coast Guard is using an IBM360/50 to predict oil spill movements in the open sea, which will enable the Coast Guard to mobilise its clean-up resources before the oil hits the beaches, and to enable the perpetrator to be traced if there is any doubt about who it was. (CW 13/4/72 p14)
Marconi completes first phase MARS: Phase one of a £750,000 computer-based message switching system called MARS (Marconi Automatic Relay System), which will link the Meteorological Centre at Bracknell with other major world centres, has been completed by Marconi Communication Systems Ltd for the Ministry of Defence. MARS links Bracknell as a Regional Telecommunications Hub to the main trunk circuit of the World Weather Watch network first planned in 1967. Each centre on the circuit has responsibility for collecting, collating and re-transmitting weather information from its own area to the nearest centres on the circuit, and will receive information from them in turn. Bracknell is linked to the World Meteorological Centre in Washington on the West, and to Regional Telecommunications Hubs in Paris and Offenbach, near Frankfurt, on the East. Moscow and Melbourne are the other two World Centres, and other hubs are situated in Prague, Cairo, New Delhi and Tokyo. Bracknell will collect weather information on 50 bps lines from outstations in an area including the United Kingdom, Ireland, Iceland, Greenland, Gibraltar, the Netherlands, and four Ocean Weather Stations as well as from merchant shipping in the Eastern Atlantic. (CW 20/4/72 p20)
CASE launches TV terminal: A computer terminal that can display sales graphs, share price charts, process flow diagrams and a wide range of graphical data that is handled in commerce, finance and engineering has been announced by Computer And Systems Engineering. Called the CASE Computek 300, the system can display 1,050 alphanumeric characters with an unlimited number of vectors on a 256 by 256 grid which permits the drawing of straight lines and smooth curves on any part of the screen. A feature of the system is its use in conjunction with television. It can be used to drive television monitor screens as display repeaters, and it can also accept picture signals from TV cameras or video cassettes for simultaneous display with the internally generated lines and characters. A direct access memory technique allows the display to be refreshed 50 times a second and dispenses with the storage tube often used in similar systems. The system gives a very bright display and allows individual vectors and characters to be updated without erasing the rest of the screen. (CW 27/4/72 p18)
GEC launches larger real time machine: Following the introduction of the GEC 2050 at last year’s Datafair, GEC-Elliott Automation has chosen its “World of Automation” exhibition which begins today, Thursday, at Borehamwood as the occasion for the announcement of the second in its current range of real time computers, the larger GEC 4080. The new machine, which is designed for medium to large industrial and military automation and control applications and features a high level assembly language called Babbage, was developed by GEC-Elliott Automation’s subsidiary, GEC Computers. The GEC 4080 can have from one to four processors, uses bipolar MSI technology, and is capable of speeds approaching one million operations a second. The processors are designed to operate on a wide range of data types, from byte-oriented communications data to 64-hit floating point formats, and have a built-in multiplexer providing up to 256 autonomous I/O channels. (CW 4/5/72 p20)
IBM mark reader for 370s and System 3: Another piece of equipment, this time an optical mark reader, designed for use with 370 machines and System 3, but which cannot be used in 360 configurations, has been introduced by IBM. The unit, the 3881, handles documents from three inches square to 9×12 inches at speeds up to 6,000 documents an hour, double the reading speed of its predecessors. The 3881 was launched at Hanover Fair and first deliveries from IBM’s Greenock, Scotland, factory are scheduled to start by the middle of next year. Price of the unit will start from £27,000, and rental on a 24-month contract will be £566 a month. The 3881 supersedes two models, the 1232 off-line unit which outputs to a card punch, and the 1231 on-line model, which has versions for 1130 and 1400 machines and for the 360. (CW 11/5/72 p1)
Brain disease diagnosis eased with EMI machine: An important breakthrough in the investigation of brain diseases has been achieved by EMI Electronics and Industrial Operations, using a number of X-ray detectors linked to a minicomputer which digitises their readings and produces a composite picture of a “slice” of a patient’s brain only about one cm thick. The machine rotates round a patient’s head taking 56,000 readings from a narrow beam of X-rays passing through the head in a single plane. The readings are digitised and the minicomputer solves 28,000 simultaneous equations and builds the results up in the form of a 160 by 160 picture matrix. (CW 18/5/72 p7)
Reader will handle all types of Cards: A card processing system designed to read any type of punched card has been introduced by Contemporary Electronics Ltd, of Aldershot, Hampshire. Available as optional extras, to work on-line to the card reader, are a small line printer, magnetic tape unit and a modem interfacing unit. Based on photoelectric techniques, the card reader forms the basis of the system which incorporates vacuum card feeding in which card wear and tear is reduced by controlled-tension card pressure rollers. The reading speed is 400 cpm. The reader, which sells at £998, will handle all punched or blackmarked cards, including 80-column cards, plastic credit cards, IBM System 3 96-column cards, and any other data carrying card which does not exceed the dimensions of the standard 80-column card. (CW 25/5/72 p1)
Face-to-face meetings at the BCS in London have now been resumed. It is as yet uncertain when Manchester meetings will recommence.
However, in view of the success of the online Zoom meetings that have taken place over the last year in allowing many more members to attend, we are now operating in hybrid mode by making meetings available over Zoom as well as in person.
London meetings take place at the BCS — 25 Copthall Avenue Moorgate EC2R 7BP starting at 14:30. The venue is near the corner of Copthall Avenue and London Wall, a three minute walk from Moorgate Station and five from Bank.
You are strongly advised to use the BCS event booking service to reserve an in-person place at CCS London seminars in case the meeting is fully subscribed. Web links can be found at www.computerconservationsociety.org/lecture.htm . The service should also be used for remote attendance.
For queries about meetings please contact Roger Johnson at
Details are subject to change. Members wishing to attend any meeting are advised to check the events page on the Society website.
MuseumsDo check for Covid-related restrictions on the individual museum websites.
SIM : Demonstrations of the replica Small-Scale Experimental Machine at the Science and Industry Museum in Manchester are run every Wednesday, Thursday and Friday between 10:30 and 13:30. Admission is free. See www.scienceandindustrymuseum.org.uk for more details.
Bletchley Park : daily. Exhibition of wartime code-breaking equipment and procedures, plus tours of the wartime buildings. Go to www.bletchleypark.org.uk to check details of times, admission charges and special events.
The National Museum of Computing Normally open Tue-Sun 10:30-17.00 but at present opening days are somewhat irregular so see www.tnmoc.org/days-open for current position Situated on the Bletchley Park campus, TNMoC covers the development of computing from the “rebuilt” Turing Bombe and Colossus codebreaking machines via the Harwell Decatron (the world’s oldest working computer) to the present day. From ICL mainframes to hand-held computers.
Please note that TNMoC is independent of Bletchley Park Trust and there is a separate admission charge. Visitors do not need to visit Bletchley Park Trust to visit TNMoC. See www.tnmoc.org for more details.
Science Museum : There is an excellent display of computing and mathematics machines on the second floor. The Information Age gallery explores “Six Networks which Changed the World” and includes a CDC 6600 computer and its Russian equivalent, the BESM-6 as well as Pilot ACE, arguably the world’s third oldest surviving computer.
The Mathematics Gallery has the Elliott 401 and the Julius Totalisator, both of which were the subjects of CCS projects in years past, and much else besides.
Other galleries include displays of ICT card-sorters and Cray supercomputers. Admission is free. See www.sciencemuseum.org.uk for more details.
Other Museums : At www.computerconservationsociety.org/museums.htm can be found brief descriptions of various UK computing museums which may be of interest to members.
North West Group contact details
Computer Conservation Society
Aims and Objectives
The Computer Conservation Society (CCS) is a co-operative venture between BCS, The Chartered Institute for IT; the Science Museum of London; and the Museum of Science and Industry (MSI) in Manchester.
The CCS was constituted in September 1989 as a Specialist Group of the British Computer Society (BCS). It thus is covered by the Royal Charter and charitable status of BCS.
The objects of the Computer Conservation Society (“Society”) are:
Membership is open to anyone interested in computer conservation and the history of computing.
The CCS is funded and supported by a grant from BCS and donations. Some charges may be made for publications and attendance at seminars and conferences.
There are a number of active Projects on specific computer restorations and early computer technologies and software. Younger people are especially encouraged to take part in order to achieve skills transfer.
The CCS also enjoys a close relationship with the National Museum of Computing.