Pages: pp. 63-69
On 8 December 2005, in the Fellows Room of the London Science Museum, Philippe Denoyelle and Hans Pufal (see Figure 1,)—both from Grenoble, France—recounted for two hours their abridged history of computing in France.
Figure 1 (a) Philippe Denoyelle and (b) Hans Pufal spoke at the meeting.
The first part of the presentation dealt with the era of calculating machines and punched cards, which included Pascal (1642), the Arithmometer of Thomas de Colmar (1820–1860), and de Prony's tables of logarithms (1800), versions of which the speakers reported as still being in regular use by the French army as late as 1940.
They even mentioned Babbage because of a visit he made to France and his self-proclaimed interest in Jean-Marie Jacquard's clever combination of the ideas of Bouchon (1725), Falcon (1728), and Vaucanson (1745) to make the Jacquard loom. Despite having assumed a great importance in the minds of historians of computing, Jacquard's automatic loom enjoyed only limited success; the real commercial breakthrough came only after significant further development, particularly by Breton (1817).
The development of the punched-card-controlled loom led naturally to discussions of punched cards, with Hollerith looming large. In 1889, Hollerith brought a copy of his prototype machine to Paris for the Universal Exhibition (the Eiffel Tower exhibition). Receiving a gold medal for his Electrical Tabulating System, Hollerith left the machine in Paris (the exact reasons are not recorded, possibly because of the prohibitive costs of returning it) and the earliest surviving example of a Hollerith machine is carefully preserved in the reserves of the Conservatoire National des Arts et Métiers in Paris. 1
The establishment, in Paris in 1914, of the International Time Recording Company marked the start of an information processing industry in France. Some eight years later, the Societé Anonyme de Machines a Statistiques (SAMAS) was established in Paris to market the machines of James Powers. The French company was so successful that its name was incorporated into the European company, Power-SAMAS, which was created in 1932.
Pufal and Deyonelle brought to a close their discussion of the mechanical era by recounting the story of Bull. The company, founded in Paris by Knut-Andreas Knutsen in 1932, was originally called Egli-Bull. Knutsen had named the company after his friend and collaborator, Frederik Rosing Bull who had developed an alternative to the Hollerith tabulator but died before being able to realize its full commercial potential. Ironically, despite the fact that the company named in his memory became a major player in the French information technology scene, Bull himself, as far as we know, never set foot in France.
Following a break for questions, during which the resemblance of the image of James Powers and the young General Charles De Gaulle was noted, the second part of the presentation moved on to a discussion of the electronic era.
Pufal and Deyonelle began by giving some much-deserved attention to the work of Louis Couffignal. Although generally neglected in France today, Couffignal wrote about automatic calculating machines as early as 1930 and advocated the use of the binary number system in a paper dated 1938. His proposition for an automatic electronic calculator in 1947 led to a patent in January 1951. Construction was started but never completed and the remains of the project lie almost forgotten in the Musée des Arts et Métiers. The speakers concluded, not unreasonably, by observing that Couffignal's work merits further research.
A more pragmatic and successful early pioneer was Francois-Henri Raymond, who in 1947 created his company the Societé d'Electronique et d'Automatisme (SEA) to pursue his ideas of developing electronic calculating systems. 2 He started out building a successful series of analog computers. In 1951, SEA delivered the Calculatrice Universelle Binaire de l'Armement (CUBE), the first all-French computer, to the military.
Little is known of this computer but recently uncovered documents seem to indicate that it resembled the EDSAC. A series of mainframe computers followed with modest success in the very limited French postwar market. In the late 1950s, SEA developed the Calculatrices Aritmétiques Binaires (CAB) 500, a small "personal" computer about the size of a desk, based on magnetic core logic and using a drum as main memory. A user-friendly programming language called Programmation Automatique de Formules (PAF) was developed, which bore a close resemblance to Basic, still some years in the future. The keywords of PAF could be specified in a number of languages.
In the early 1950s, the company Bull was successful in building electronic calculating engines to attach to, and increase the performance of, its punched card tabulators. The Gamma 3 calculators were developed over a period of years and the connection of a drum memory converted them into real computers. The introduction of electronics at the low end of the market caused perturbations at IBM, which was working on large-scale electronic computers at the time. High-level discussions at IBM on the response to the Gamma 3 eventually led to the introduction of the 1401 small-scale transistorized computer in 1959. The early conceptual designs of what would become the 1401 were developed by a joint Franco-German team of engineers working out of the IBM European laboratories in Paris and Stuttgart.
In the late 1950s, Bull embarked on building a large computer, the Bull Gamma 60. This machine had an innovative architecture boasting multiprocessing and multiprogramming features not unlike those found in some modern architectures. Software for the Gamma 60 was an afterthought and while approximately 20 systems were eventually delivered, the project was not a financial success. Sapped by the cost of the Gamma 60 development and faced with competition in the form of IBM's successful and less expensive rival, the 1401, Bull had no answer and the very survival of the company came into question.
Unsuccessful attempts were made to secure support from the French government, with the result that General Electric took over the Bull company in 1964. A series of events in the mid-1960s, including the refusal by the US government to sell a supercomputer for French nuclear research, led the government to conceive of the "Plan Calcul"—an ultimately ill-fated attempt at creating an independent French national computer company. The forced merging of a number of small companies into the new entity called Companie Internationale pour Informatique (CII) and the specification of a product line by central government sealed the fate of the initiative which, courtesy of government support, survived for around 10 years before being sold to Honeywell-Bull for the token sum of 1 French franc.
The year 1973 marked the advent of the microprocessor. In that year, the French computer company Réalisation d'Etudes Electronique (R2E) made history by delivering the world's first microprocessor-based computer. Constructed around the then recently announced Intel 8008 microprocessor, the Micral-N, was dubbed a microcomputer in an article in Byte magazine's June 1973 issue. This was the first printed use of the term microcomputer. By 1980, a number of other actors had entered the market and France experienced the microcomputer boom. The government attempted to help the industry with an ambitious plan to provide all schools with microcomputers. In the end, the project failed and only half of the original 120,000 computers planned were ordered. Faced with ever-growing dominance of PC compatibles, most French companies tried to compete by developing compatible models but none survived.
The story of the French software industry probably merits a complete seminar. In this talk, the speakers concentrated on two early aspects: the emergence of a university research laboratory in Grenoble and the subsequent creation of Société pour la Gestion de l'Entreprise et le Traitement de l'Information (SOGETI).
In the late 1940s, Jean Kuntzmann, a professor of mathematics, arrived in Grenoble—a relatively small university town that over the previous century or so had developed hydraulic, hydroelectric, and electrochemical industries. The region had developed a strong culture of cooperation between the industrial and academic worlds and Kuntzmann was asked to develop mathematical tools to solve apparently intractable problems facing the industry. The concept of applied mathematics was foreign to the French establishment at the time and it was only his energy, determination, and stubbornness that permitted Kuntzmann to finally establish his laboratoire de calcul.
Working at first with mechanical hand calculators, the team quickly showed its value and in 1952 installed an SEA analog computer, followed in 1958 by a Gamma 3 computer. By 1960, Kuntzmann set Grenoble firmly on the map of computing by establishing a course in computing and applied mathematics, the first in France. Industry followed the lead, with IBM establishing an important research center near Grenoble and installing a 360–67 computer at the university (in full contradiction to the Plan Calcul). Several Algol compilers were developed and then came the first IBM PL/1 compiler. Today, Kuntzmann's legacy lives on in the Institut d'Informatique et de Mathématiques Appliqués de Grenoble (IMAG), which groups eight world-class teams specializing in computer science and applied mathematics.
In 1967, a graduate of Kuntzmann, Serge Kampf, recognizing that the demand for expert information technology services was destined for explosive growth, established SOGETI in Grenoble. The company went from strength to strength and after many mergers and name changes survives today as Capgemini. The company is one of the top five leading management consulting and outsourcing service companies in the world.
The last part of the presentation introduced the principal actors working in France to preserve the computing heritage. ACONIT (the Association for an Academy of Data Processing and Telematics), is a small association in Grenoble with a large collection of computing artifacts. ACONIT celebrated its 20th anniversary this year. The group was created in 1985 by a group of academics and industrialists who were dismayed by the rapid and definitive disappearance of computing history. Their primary goal was the creation of a conservatory of information technology where, in analog to its music counterpart, computing technology and its history could be studied, conserved, and transmitted to future generations.
The Cité des Sciences et de l'Industrie, the Paris equivalent of the London Science Museum, is emphasizing its pedagogical mission over heritage conservation. The Conservatoire National des Arts et Métiers, created in 1794, and its museum arm, the Musée des Arts et Métiers, are the principal state institutions concerned with the preservation of technical and scientific heritage. Recently, a major project was launched to conduct an inventory of the surviving technical and scientific artifacts from the past half century and to develop a Web-accessible database of that inventory.
The company Bull actively supports an association of retired employees who are charged with the conservation of the rich history of the group. The Fédération des Equipes Bull (FEB) manage a collection of hardware, which includes one of almost all products built by Bull. A small-but-active group of retired engineers restore the old machines to full working state. The FEB succeeded in having a 1932 Bull tabulating machine classified as a French historic monument, the first technical artifact to be so categorized. Subsequently, ACONIT also received the classification of a 1950s SEA analog computer as a French historic monument.
A list of private computer collectors would be quite long, but at the top of the speaker's list would be Stephan Mathon, an ex-IBM account executive who in the 1970s decided to preserve the machines his clients were systematically discarding. He personally bought the machines he was replacing and has built an impressive and coherent collection of mainly IBM mainframes of the second and third generation.
In conclusion, recent and not-so-recent publications on the history of computing in France were cited. Computer history conferences are organized every two or three years in France and proceedings of the first two, dating back to 1988, were presented to the Computer Conservation Society for their library.
Philippe Denoyelle and Hans Pufal are active members of ACONIT. Denoyelle is retired after 35 years in the French IT industry as engineer and manager. Pufal is changing career paths after 30 years as a professional in software engineering and is attempting to create a new profession: paleoinformatics.
In an afterword, the speakers made a plea for the identification of a mystery object (see Figure 2). Long held in the ACONIT collection, marked ICT, France, it is clearly related to punched cards but seems to punch three columns of 24 lines of small round holes. The audience was unable to make any firm identification but the question created some interest and several suggestions were made as to where to find more information. Perhaps readers of Annals may be able to identify this artifact?
Figure 2 Different views of the mystery object.
The Information Processing Society of Japan (IPSJ) is celebrating the 50th anniversary of the first Japanese electronic digital computer, called FUJIC (Fuji Computer). The society held an anniversary exhibition from 7–9 March 2006 during its 45th annual convention at Kogakuin University in Shinjuku, Tokyo. About 2,300 people participated. Deborah Cooper, the 2006 IEEE Computer Society President, offered congratulations to Professor Yuichiro Anzai, President of IPSJ as well as Keio University, on the 50th anniversary.
The exhibit included a slideshow of major Japanese computers played on six large display monitors for each decade—from the 1950s to the new millenium (see Figure 3). In the 1950s and early 1960s, relay computers, vacuum tube computers, parametron computers, and transistor computers were developed almost in parallel in Japan. (Note: Parametron, invented by Eiichi Goto in 1954, is a majority logic element utilizing parametric excitation.) This situation and the later developments of mainframes, supercomputers, office computers, minicomputers, and personal computers were shown in a time line on a large, long board (see Figure 4).
Figure 3 Slideshow of major Japanese computers over the past six decades.
Figure 4 Large, detailed time line of computer developments in Japan.
The primary computer being celebrated—the FUJIC—was a vacuum tube computer built by Fuji Photo Film Company in March 1956. Its full-sized picture was displayed, as the original is a part of the permanent exhibition at the National Science Museum in Ueno, Tokyo. The picture was just like looking at the real machine, as Figure 5 shows.
Figure 5 Full-sized image of the 1956 FUJIC vacuum tube computer, developed by Bunji Okazaki of Fuji Photo Film Company.
The FUJIC was developed by Bunji Okazaki of Fuji Photo Film Company to automate the calculation of the lens design process. Okazaki started to research building a computer in 1949 and completed the FUJIC in March 1956. Surprisingly, during those seven years, he did most of the investigation, design, and manufacturing work by himself.
The FUJIC consisted of logic, memory, input, and output units. About 1,700 vacuum tubes were used for logic circuits. Mercury delay lines were used for the memory unit with 255 words of 33 bits each. Okazaki even manufactured an optical card reader for the input unit.
Only three vacuum tube computers were built in Japan—FUJIC, the Tokyo Automatic Computer (TAC) from the University of Tokyo, and Osaka University's computer. The design of the latter two was based on the EDSAC architecture but Okazaki designed his original one for the FUJIC. It is a binary three-address machine with an accumulator also available to store a calculation result for the next operation. This is how it effectively shortened lens design calculation programs.
Another unique feature of the FUJIC was that Okazaki controlled the frequency of the memory clock rate instead of the mercury temperature to keep the same number of bits stored in a delay line. The clock rate for logic circuits was 30 kHz, but arithmetic operation speed was fast enough for lens design calculations, as it employed parallel arithmetic operations. Okazaki's had originally wanted it to be 1,000 times faster than that of manual calculations but it surpassed that and realized a speed of about 2,000 times. After its completion, the FUJIC was used not only for lens design calculations for the company, but also for various calculations requested by other organizations.
Also on display was an original ETL Mark IVA transistor general-purpose computer developed in 1959 and an original Japan National Railway's MARS 101, which was an online seat reservation system developed in 1964 (see Figure 6).
Figure 6 The original ETL Mark IVA (on the right, developed in 1959) and the original MARS 101 (on the left, developed in 1964).
The ETL Mark IVA was developed by Electrotechnical Laboratory (ETL). This is an enhanced version of the ETL Mark IV developed by ETL in 1957. Both are decimal transistor computers using junction-type transistors for logic circuits. The ETL Mark IVA introduced magnetic core memory. The ETL Mark IV used a magnetic drum memory. Before building the ETL Mark IV, the ETL Mark III, the first transistor computer in Japan, was developed as early as July 1956. It was a prototype of a transistor computer using point-contact type transistors for logic circuits and solid glass delay lines for memory. These ETL machines adopted dynamic flip-flop circuits to minimize the number of transistors in use. NEC, Hitachi, Matsushita, and Hokushin-denki built commercial products based on the ETL Mark IV architecture and circuits.
The Multi-Access Seat Reservation System (MARS) 101 was an online, real-time seat reservation system designed by JNR Railway Technical Research Institute and was manufactured by Hitachi. It was used from February 1964 until 1971 all over Japan and is well-known as the Green Window of JNR. MARS 101 was an advanced task-sharing multicomputer system; it consisted of a main computer, a communication computer, a table search computer, and a seat file computer. The main computer controls the subcomputers to make the tasks flow optimally and to obtain maximum throughput from the system. This became the first online real-time, large-scale computer system in Japan. MARS-1, the prototype of MARS 101, was the first online real-time seat reservation system for any railway train in the world and began its service in January 1960.
Readers can see more historical Japanese computer pictures—including the FUJIC, ETL Mark IVA, and MARS-101—on IPSJ's Computer Museum homepage: http://www.ipsj.or.jp/katsudou/museum/index_e.html.Akihiko YamadaTokyo Denki Universitya.email@example.com
James E. Tomayko, a professor of software engineering at Carnegie Mellon University and historian of technology, passed away on 9 January 2006 at the age of 56 after a long illness. In memory of his long involvement with the Annals editorial board, we will publish an article in the Biographies department in an upcoming issue.Chigusa KitaKansai Universityannalsfirstname.lastname@example.org
The Computer History Museum ushered in the new year with a high-energy panel discussion on 11 January 2006 by the four original founders of Sun Microsystems—Andy Bechtolsheim, Bill Joy, Vinod Khosla, and Scott McNealy.
On 18 January the retired board chairman of EMC—Michael Ruettgers—shared personal stories from his multidecade odyssey in the high-tech industry, and on 14 February a panel discussed social computing from early bulletin board system (BBS) systems to blogs.
A truly unique panel discussion took place on 27 February, during which four original employees of Shockley Semiconductor Laboratories—arguably the birthplace of Silicon Valley—discussed their early work with William Shockley, co-discoverer (and Nobel Prize winner) of the transistor effect in 1947. The panel was moderated by historian of physics Michael Riordan.
The day before the panel discussion, these four innovators and about 50 other Shockley employees, their families, and supporters met at the original location of the labs, now the wholesale fruit market at 391 San Antonio Road in Mountain View, California (hometown of the museum). Still poking through the ceiling over the neatly arranged tables of avocados and pears are holes from when Shockley Labs had its fume hoods exhausting toxic gases through them.
Additionally, senior curator Dag Spicer conducted a series of eight mini oral histories with Shockley employees who attended the event. The Computer History Museum saw continued progress on its usual activities of volunteer days (with a pool of nearly 250 active volunteers) as well as with its oral history program. Oral histories conducted since the last issue of Annals include George Michael, Harry Huskey, Fernando Corbato, Gordon Bell, Alan Shugart, Ivan Sutherland, Paul Baran, and a dozen others.
Additionally, there are three new projects to report:
The museum is also holding a gala event on 15 May at which its newly restored PDP-1 computer will be publicly unveiled. This is the result of two years of effort by a team of dedicated volunteers. Those involved drew on the world's largest archive of DEC-related materials (which resides at the museum).
A software workshop titled, "The Attic and the Parlor" is to be held 5 May. This one-day, invitation-only event will attract scholars who work with the issues attendant on software preservation from around the world. Results of the workshop will aid in shaping the museum's long-term software preservation and access policy.
All lectures, panel discussions, and oral histories noted in this summary are available online as media streams. Please see the museum's Web site: http://computerhistory.org.Dag SpicerComputer History Museumspicer@computerhistory.orgBlocks Make Difference Engine
If you visit http://acarol.woz.org/, this Web page documents Andrew Carol's model of a mechanical difference engine. Incredibly, this machine, which operates up to two orders of differences on three-digit numbers, is constructed entirely from standard Lego Technic construction set parts.
The page is headed by a picture of the model, together with a brief history of Charles Babbage's difference engines (which provided the inspiration). Next, a simple example of using the method of differences to tabulate a polynomial is given. More in-depth Web resources, and further high-resolution pictures of the model are available via hyperlinks from the main page.
Carol gives a fairly detailed description of the mechanism, offering fascinating insights into the challenges that he had to overcome to build this model entirely within the confines of the Lego system. The model pictured is the result of dozens of design iterations, which steadily refined the operation of the various mechanisms. The design is modular and can be broken down into individual units and reassembled in a matter of minutes.
Carol says he is working on expanding the machine to three orders of differences on four-digit numbers and simplifying the construction. He promises a major update to the Web page in the near future, with more detailed descriptions and pictures of the principal subassemblies. His aim is to make it possible for others to copy his work.
Unfortunately, a Web page cannot do full justice to the great skill of the model's creator in finding elegant solutions within the Lego system to the many technical issues encountered. Therefore, on 17 February 2005, Carol brought the machine to the Computer History Museum in Mountain View, California, where he demonstrated it and explained the intricacies to an audience of museum staff and docents. It performed flawlessly, calculating a table of squares. However, at approximately 150 turns of the crank for each result produced, it may win a prize for the slowest calculator ever!Tim Robinsontbr00@pacbell.netRetirement of Reader's Digest Database
Baseline magazine reported the demise of the Unified File System at Reader's Digest Company, a marketing database that had been at the center of the firm's operations for 35 years. Originally designed by a programming team of 30, the database ran on an IBM System/360 and grew to have more records than the US had households. The company began to consider how it might replace the program in 1990 and completed the job in May 2005. Baseline gives a full history of the system at http://www.baselinemag.com/article2/0,1540,1835264,00.asp.Anne DonelanIDeadonelan@ide.com