behind the curtain at leo: a personal reminiscence

IEEE Annals of the History of Computing Published by the IEEE Computer Society 1058-6180/03/$17.00 © 2003 IEEE 3

The 50th anniversary of business computing,with special recognition for the LyonsElectronic Office (LEO) business computer, wascelebrated in London in 2001. The anniversaryoccasion, coupled with the publication of threebooks,1-3 and two articles in Annals,4,5 meantthat LEO was emerging from obscurity at lastand achieving its rightful place in computerhistory.

The bare facts of how LEO came to be builtby a London, England, catering firm are wellknown now. Figure 1 lists key dates6 in thecomputer’s history.

The story that follows is based on my per-sonal recollections during the time I was themanager of Lyons’ Systems Research Depart-ment and then of LEO’s Systems Analysis andApplications Department, to which were lateradded Consultancy and Marketing.

The environmentThe catering company that built LEO, J.

Lyons and Co., was an old established businesswith a management structure of a type that hassince largely disappeared. Its founders weremembers of a family that had emigrated toEngland from Prussia in 1841. One of theGluckstein brothers married a Salmon, and sothe dynasty of Salmon and Gluckstein wasfounded. Their business began with tobaccoproducts. Later, they branched out and estab-lished the Lyons catering business.7

After a time, so many family members werein the business that to avoid confusion theywere all known by their first names prefixed by“Mr.,” save for the wartime chairman, who wasknown as “Sir Isidore,” and another of the oldergeneration who was addressed as “MajorMonte.” Each division and subsidiary was theresponsibility of a family member. Below themwas generally a senior manager responsible forrunning the day-to-day affairs, but strategicdecisions were always made by the family.When the computer subsidiary, LEO ComputersLtd., was set up, Lyons named Anthony Salmonto head it. Among his recent responsibilitieshad been control of the company’s tea gardensin Malawi. The family had offices in the com-pany headquarters in the west of London; con-sequently, decisions could be made rapidly.Such was the case when Lyons agreed to builda computer for the company’s own use.

Over many decades, the company haddeveloped a can-do culture. It had taken onenormous banquets calling for meticulousorganization; accordingly, the company wasconfident that the seemingly impossible couldbe achieved as long as it was properly organized.Lyons was essentially a vertical company, pre-pared to take over any activity for which therequired level of service could not be obtained

Behind the Curtain at LEO: A Personal ReminiscenceDavid Tresman Caminer

This reminiscence recalls the environment in which LEO came to bebuilt, the personalities involved, the major problems that wereencountered, and the circumstances that led to the disappearance ofthe brand at the time when it was achieving its greatest successes.From first to last, LEO’s active life span was only 20 years.

1949—Decision made to design and build an electronic computer toserve the Lyons business.

1951—The world’s first business computer job starts running regularlyon the basic LEO.

1951—The world’s first full-scale business computer job, the Lyonspayroll, ran on the completed LEO l computer. LEO Computers set upas a subsidiary of J. Lyons and Co.

1962—The multiprogrammed, solid-state LEO lll was delivered threeyears ahead of its IBM equivalent.

1963—Merger of LEO Computers and English Electric computerdivision.

1969—The final deliveries of LEO 326 computers (the fastest of theLEO lll range) were made, completing 61 LEO lll deliveries in all.

Figure 1. Key milestones in LEO development.

elsewhere. So, it seemed less remarkable frominside Lyons than from outside that because thedesired equipment was not commercially avail-able, the company would naturally fill the gap.

PersonalitiesThe LEO computer and its applications

evolved from the vision and efforts of severaldifferent personalities.

Raymond ThompsonFirst and foremost was Raymond Thompson,

the company’s deputy chief comptroller sent tothe US in 1947, leading a two-man party toinvestigate developments in the office manage-ment field that had occurred during World WarII. It was that party’s report that set everythingin motion. The computer section of the reportis reproduced in LEO—The Incredible Story of TheWorld’s First Business Computer.2

TRT, as he was generally known, was aCambridge mathematician with first-class hon-ors. He had joined Lyons in 1931. JohnPinkerton, LEO’s chief engineer, said of him:

He had the quickest intellect of anyone I evermet. Sometimes when trying to persuade him ofthe value of some new approach, he would makea momentary resistance, but if you were right,then in a second or two his attitude wouldswitch, your idea was seized on and elaborated inways beyond anything that had occurred to you.8

Pinkerton made this statement from a back-ground—which included Cambridge Universi-ty and government research—that led him tohave met some of the leading brains of ourtime. Perhaps not surprisingly, TRT sometimesfound others to be somewhat slow. TRT wasnotorious for standing up two sentences into apresentation and saying, “What you mean tosay is this …” Generally, but not always, he wasright. Picking up new ideas extraordinarily rap-idly, he was sometimes in danger of adoptingthem as his own. Len Lenaerts, who amongother duties was responsible for improving theLEO system reliability and for participating indesign, suffered notably. TRT found it difficultto grasp that a pre-war clerk in his departmentcould have engineering ideas of such originali-ty. To some extent, this outcome resulted fromthe company’s rigidly hierarchical nature. LeoFantl, an early LEO programmer, commentedbitterly, “He was able to be brutally frank andhe unthinkingly made me suffer for my non-scientific background and do-it-yourself‘education.’”2 Fantl had come over fromCzechoslovakia on a Kindertransport and had

done much of his study as a farm laborer beforejoining the Royal Air Force (RAF).

On the other hand, Thompson’s enthusiasmwas an essential motor of the whole project. Hewas an enthusiast not only in his work, but inwhatever he was doing. He enjoyed teachingwhat he had learned himself and later tookclose interest in all areas of computer training.

John SimmonsJohn Simmons was Lyons’ chief comptrol-

ler, to whom Thompson presented the reportof the US trip and the recommendations stem-ming from it. Simmons was another Cam-bridge mathematician with a first-class honorsdegree. He had been recruited by Lyons in 1923when, with expansion, the company founditself engulfed by paperwork. When he arrived,the atmosphere was Dickensian, with highmahogany desks and stools. Over the years,Simmons had mechanized the offices with cal-culating and accounting machines from Bur-roughs and NCR. However, it was alwaysSimmons’ intent that the purpose of the mech-anization was not just to carry out the officetasks more rapidly and economically but,equally important, to provide managementwith better information as to why profits var-ied and how variations could be put right. Tosupport him in this work, he created a SystemsResearch Office that focused on creative organ-ization and methods.

Simmons was as reserved as Thompson wasoutgoing. He rarely visited the offices for whichhe was responsible, relying mainly on thoseimmediately reporting to him to develop andexecute his plans and to keep him informed.He was the son of two missionaries and wasguided by a strong social conscience. In arecorded interview in the 1970s, he declared:

The difference between a routine office job anda routine factory job is that in the factory job youare doing the same thing, literally the samething, time and time and time again. In theoffice you are doing the same kind of thing, butyou are feeding in different figures to themachine all the time. You have got to concen-trate. You can’t even daydream.2

His ambition was to get rid of this drudg-ery. He spoke of his reaction when the com-puter proposal was made: “They put a matchto a bonfire that was already latent. The sug-gestion only had to be made to be one ofthose exciting things that are immediatelyobvious.”2 (Figure 2 shows Simmons at a com-pany function.)

4 IEEE Annals of the History of Computing

Behind the Curtain at LEO

John PinkertonAt the operational level the computer

responsibilities were divided into two areas.John Pinkerton was responsible for buildingthe system, and I was transferred from LyonsSystems Research to form a new unit, respon-sible for making the system operational. Weboth reported to Thompson. Pinkerton hadbeen recruited from Cambridge. It had beenagreed by the board of directors that the com-puter project would go ahead, provided thatthe EDSAC system under construction atCambridge, based on the von Neumann archi-tecture, had demonstrated that it could suc-cessfully run a stored program. BeforePinkerton was hired, Lyons had neveremployed an electronic engineer. There was astrong electrical department, but its work waslargely devoted to providing and regulatingthe Lyons power supplies.

Pinkerton had spent the war years inTelecommunications Research Establishment,the prestigious UK government research organ-ization of which the principal objective was todevelop electronic aids to the war effort. Afterthe war, he had returned to Cambridge forpostgraduate studies.8 When Maurice Wilkes,the director of the mathematical laboratorythat was building EDSAC, was asked by Lyonsto interview Pinkerton, he replied thatPinkerton was just the man for the job.Pinkerton, approaching 30 when he wasappointed, had a calm temperament thatenabled him to overcome many problems dur-ing the design and building of the LEO systems.He also had to withstand pressures fromThompson, who was prone to underestimaterequired project resources. One of my abidingmemories is Pinkerton standing on the LEO Iplinth and resisting pressure to bring forwardthe forecast completion date for LEO II. “Whatare the problems that make you need so muchtime?” demanded Thompson. Pinkertonpaused, then replied quietly, “You are askingme to give you the list of the towns in Chinathat I don’t know the names of.”

My roleMy responsibility of preparing productive

applications to be ready for the equipment assoon as it reached completion was seen as anatural extension of my work as manager ofSystems Research. As I’ve commented else-where, it was like going from checkers to three-dimensional chess.3

However, I was able to approach the new taskwith a detailed knowledge of the Lyons tasksthat were to be converted. Before the war, I had

worked at Lyons as a management trainee inSimmons’ organization. The preparation couldnot have been better. During the war, I servedwith the Green Howards infantry regiment inNorth Africa. After being wounded, I wasbrought home and rejoined Lyons, where thework that I led in developing systems engineer-ing is documented elsewhere.4 Much of the suc-cess that LEO Computers achieved in producinga working system came from the close relation-ship that Pinkerton and I quickly established.

Len Lenaerts and Derek HemyThis survey of the individuals involved in

the LEO project’s earliest days would not becomplete without reference to two remarkablecharacters who emerged from Lyons’ existingresources. One was Len Lenaerts, already men-tioned, and the other was Derek Hemy, hiscounterpart on the programming side of theoperation.

Lenaerts’ involvement began with the bar-gain Lyons had struck with Wilkes that, inreturn for access to the resulting work, Lyonswould provide a sum of money and the servic-es of a technician for a year. Lenaerts was thattechnician. He had come straight to Lyonsfrom school to work as a clerk. The job was notmuch to his taste, and he made unsuccessfulefforts to escape over the years. The war pro-vided him with an opportunity to follow histechnical interests. He became a trainee wire-less mechanic in the RAF and then obtained ajob in radio countermeasures. On promotion

April–June 2003 5

Figure 2. John Simmons and David Caminer at a LEO celebration in the1950s. (Photo from David Caminer’s personal collection.)

to sergeant, he was responsible for threeLondon RAF jamming sites.

Lenaerts returned to Lyons in an unspeci-fied capacity after the war, and he was readilyavailable when Lyons needed a technician tohonor the commitment to Wilkes. His last jobbefore leaving for Cambridge was said to havebeen to build an automatic coin-in-the-slotsausage-frying machine. At Cambridge he notonly learned how EDSAC was constructed butwas able to make a positive contribution tounit circuit design.

Hemy, too, had joined Lyons pre-warstraight from school as another of Simmons’management trainees. Wartime service tookhim first to the Royal Engineers’ chemical war-fare branch and then to the Royal Signals,where he established and commanded a unitwhose principal task was radio fingerprintingto identify the source of enemy transmissionsthrough the analysis of their electronic charac-teristics and transmission styles. Back from thewar, Hemy joined me in Systems Research, andhe was a natural in learning about the pro-gramming techniques that were still in theirinfancy at Cambridge. With a young researchstudent, David Wheeler, who later became aprofessor of computer science at Cambridge, hemapped out the first, embryo approach to thepayroll application. The requirements for busi-ness programming varied considerably fromthose required for scientific programming. Boththe nature of the problems and the composi-tion of the user community were different.Hemy’s work enabled us to develop techniquessuitable for business programming, step bystep. In the Annals’ obituary about him, he wasdescribed as “invaluable.”9 Nothing could havebeen truer.

ProblemsAn important part of the work of the Lyons

computer team lay in unmapped territory, par-ticularly in regard to input and output (I/O). Inscientific computing—chiefly the only kind ofcomputing that existed up to that time—theretended to be little data, much computation,and few results. The type of ballistic table com-putation for which ENIAC was built was repre-sentative. Little data was needed for eachballistic trajectory, but many repetitive calcula-tions were needed to track the projectile to itszenith and then more calculations as the shellcame to Earth. Typically, the only resultsrequired were the definition in distance anddirection of the landing point. The next set ofdata, giving perhaps a slightly raised elevation,could then be applied automatically by the pro-

gram. Such was the character of a great deal ofcontemporary scientific work, and it differedlittle from the work that Charles Babbage hadlabored on in the previous century. For theseneeds, existing input and output mechanismswere adequate.

For business needs, it seemed evident thatunless some way could be devised to feed indata and record the results more rapidly, thespeed of the electronic computer would besquandered. From the outset in the Thompson-Standingford report2 in 1947, it was acceptedthat this medium must be magnetic wire, or,more likely, magnetic tape, which was then stillunder development. In their report they wrote:“As an alternative to the magnetic wire, a papertape, coated with a magnetic layer, has beendeveloped. It is less expensive than the wireand can be run at one tenth the speed to givethe same output.”2

It was assumed—although no classicalSystems Office Research study was carried out—that magnetic wire or tape would be the I/Omedium and, for that matter, the medium ofsecondary storage. The LEO team’s assumption,untested in any way, was that electronic com-puting required magnetic services.

Pinkerton began work on magnetic storagesoon after he joined Lyons. Magnetic tapeexpertise was not readily available, but, as ithappened, a leading electrical firm, StandardTelephones and Cables (STC), was theninstalling a new telephone exchange at Lyonsheadquarters facility, and their subsidiary,Standard Telephone Laboratories (STL), wasexperimenting with magnetic tape in conjunc-tion with teleprinters.

In May 1949, the same month as EDSACproduced its first results, STC was given anoutline of what LEO Computers envisioned.Soon afterward, a study contract was negoti-ated and a contract to build the requiredequipment followed. The plan was to recordthe data onto the magnetic tape using a spe-cial device incorporating a check function,and then feed the data into what was called aconverter, where the data would be translatedfrom its decimal notation on tape to the bina-ry notation used by the computer. Then,when the arithmetic unit was ready for eachset of data, it would be automatically flashedover to the main computer memory. Outputwould be handled similarly. Thus there wouldbe something like a three-ring circus, withresults for one unit being output while calcu-lations for the next unit were being carriedout, and the input for a third unit being readin (see Figure 3).



The plan failed to materialize. Themagnetic tape encountered stop/startproblems, but the major mistake wasour relying on system logic in a tech-nologically new device, the multi-cathode tube. These vacuum tubeshad speed advantages, but continual-ly disrupted trials by their pronenessto faults, including the appearance ofcharges on more than one cathode ata time, which was theoreticallyimpossible.

After a frustrating period, it wasclear to the Lyons team that the STCequipment would not be ready in asufficiently reliable state at anythinglike the date in 1951 of Pinkerton’scompletion of the mainframe. At ameeting, Simmons, Thompson,Pinkerton, Pinkerton’s deputy, ErnestKaye, and I in October 19516 decidedto design and build an alternativesystem as a fallback to the STC sys-tem on which we would continue tocooperate. After a time, though, theSTC project was abandoned, and allLEO resources were concentrated onthe in-house system.

In 1953 the in-house system wascomplete. It was based on the same architectureas the system worked out with STL but accom-plished its ends using conventional equip-ment—paper tape for recording and checkingdata, paper tape transports and punched cardsfor feeding data, and a line printer and card

punches for output (see Figure 4). Equallyimportant, the logical circuits were controlledby established vacuum tubes.

Although the new configuration was not asfast as the magnetic tape equipment on the STLsystem had promised to be, it was still capable

April–June 2003 7

Figure 3. LEO concurrency. Data is input for item 3 at the same time as the computations are takingplace for item 2 and results are being output for item 1. (Sketch by David Caminer.)

Figure 4. LEO I schematic. The three data channels feed data into electronicbuffers where they are ready for use by the arithmetic units. Output was handledsimilarly. (Sketch by David Caminer.)

of running the benchmark Lyons payroll job in1953–1954 twice as fast as originally hoped for.This was due to extremely fine-tuned systemsanalysis and programming.

Simmons commented afterward about thechoice of magnetic tape:

It was a decision that was to produce moreheadaches than the whole of the rest of the proj-ect put together. If we had been content to usealready existing systems for providing input andoutput we would have saved at least a couple ofyears and a great many headaches.2

The mistake meant that LEO I would not becomplete until the autumn of 1953. However,work had proceeded on the Lyons payroll appli-cation in parallel, and trials were carried out aseach aspect of the configuration was completed.

The specification10 of the resultant applica-tion—the world’s first fully integrated businessjob—reveals the detail to which the systemsanalysis was taken. The timing of the job, how-ever, depended on all three channels of data,the computations, and the two output chan-nels operating concurrently. This required agreat deal of thought. For example, only oneprint cycle could be afforded for the employeepay slip, but this would have resulted in a longthin strip that would have been difficult tohandle. Instead it was decided to give eachemployee two print cycles but to have two suc-cessive pay slips printed side by side. Thismeant storing one of the sets of pay slip details,but it was judged that in this case the addi-tional use of precious memory was justified.Counting in all preparatory runs, the overalltime per employee was 1.5 seconds:11 this on amachine that had been described as having ahundred-thousandth of the power and storageof a modern PC.

An important feature was the operation’scomprehensiveness. It was some years beforecomputers came to be regarded elsewhere as anacceptable vehicle for payroll, but even whenthey were, it was unusual for the application tostart at the clock card and go all the waythrough to the pay envelope. In the LEO case,there was just one exclusion from the comput-er calculations—the premium bonus pay-ment—which required calculating thepayment for groups rather than just for indi-viduals. There was no problem making the cal-culation by hand and passing the amount foreach employee to the computer for incorpora-tion in the gross pay.

While aiming at integration, it became amatter of policy in LEO applications never to

endanger the reliability or timeliness of a rou-tine to feed the vanity of completeness. Manymajor applications elsewhere have founderedover the years by the resolution to leave noth-ing out, whatever the economic and securityconsiderations.

The backupThe Lyons payroll application was checked

out by pilot and parallel runs and pronouncedready for service by the applications team inthe autumn of 1953. However, because therewas still no backup machine, Simmons wasreluctant to run the slightest risks and so thetrials went on until the system went live earlyin 1954. As he said later, “[F]rom then onwardswe kind of lived on tenterhooks, afraid that themachine might break down and that theBakery staff wouldn’t get their money.” Hewent on, “but in fact it never did break downand by the end of the year we had got aboutten thousand people on the payroll.”2

The lost two years had not been wasted. Assoon as it had become clear that there would bea severe delay in commissioning the full sys-tem, I specified a management accounting jobthat required little more than the facilities pro-vided by the basic EDSAC. Known as the bak-ery valuations job, its implementation has beendescribed in LEO—The Incredible Story.2 It ranfor the first time in November 1951 and con-tinued week by week for several years. With hisstrictly logical mind, Simmons had been loathto permit precious machine time to be used ona job that, as he saw it, could have been carriedout on the scientific machine at Cambridge. Inpractice, the job became a valuable testbed forthe systems analysis and programming tech-niques that had been worked out. The teamlearned, for example, how careful the precau-tions needed to be to prevent a rotten apple inthe data from corrupting the barrel.

Service bureau activityIn this period, too, Lyons performed a great

deal of scientific and actuarial work on a serv-ice bureau basis. None of this work was sought,but as soon as the success of the bakery valua-tions job became known, there was a steadystream of mathematicians, crystallographers,meteorologists, and representatives of insur-ance and aircraft companies wanting to use thisnew facility. Organizations were charged at astandard rate, but at the outset, we freely assist-ed them in getting their jobs operational.Generally, LEO Computers did not appropri-ately charge for this, so if the operational runswere short, the whole effort could be uneco-



nomical. Later, at the completion of the Lyonspayroll job, a formal operating section wasinstituted under Tony Barnes, who had joinedus after service as a lieutenant instructor in theRoyal Navy, and the arrangements becamemore businesslike. From that time, the idea ofa service bureau as a profit center developed.

The service bureau work was characterizedby highly repetitive calculation loops; it pro-vided good practice for the programmers in car-rying out optimization. Saving an instructionbecame a favorite pastime, and this skill becamea valuable asset in business programming, too.A second feature of this work was that runs wereoften long, and it was important that if a breakoccurred, as little runtime as possible should belost. To achieve this result, we devised a systemof routine checkpoints and restarts.

In a sense, the mathematical programs pro-vided a real-life set of test programs to comple-ment the comprehensive set of standard teststhat the engineers and programmers haddevised for the system. The test programs weretools in the determined efforts over this periodto improve the system reliability and makerepairs more rapidly.

It was a second major difference from scien-tific work that in business, producing timelyresults was as essential as producing accurateresults. In most computer work of the period,scientific work was being executed so muchfaster than before and such a load was beingtaken off scientists that it mattered little if theresults were a day or two late. That was a lee-way that was seldom available for business. Itwas not surprising that Simmons was con-cerned about the bakery operatives’ paychecks.

LEO IIWithout any formal decisions to the con-

trary, the initial intention that live work wouldalways be covered by a backup machine hadevaporated. As a token of this delay, the size ofthe Lyons payroll handled by the computer waslimited to 10,000 personnel until the backupwas online. It was not until July 1954 thatplans were finalized for a LEO II (see Figure 5).The original machine was given a new namebecause the engineering had been furtherdeveloped. The temptation to take advantageof experience and technological advancementsproved stronger than the anxiety to produce abackup. Among other changes the storage sys-tem was refined and a new, smaller, faster mer-cury delay line system was incorporated. Thewhole configuration was faster and, because ofthe reduction in memory tube length, less dif-ficult to keep in perfect synchronization.

A second factor in the delay was the compa-ny’s decision to build on the payroll success byoffering LEO systems for sale. New pressureswere encountered from some potential cus-tomers who wanted features they might haveread about elsewhere to be included in what-ever configuration they ordered from Lyons.

It was not until May 1957 that LEO II wasready for service. It was something of a miraclethat Lyons had been spared a calamity.Although the Lyons payroll had been limited to10,000, the payroll for a LEO customer—theFord Motor Company car plant in Dagenham,England—had also reached 10,000, and therewere other smaller payrolls enjoying the service.Additionally, the delivery service to the 200Lyons tea shops was being controlled by thecomputer, together with the paperwork for thedistribution to grocery stores all over the coun-try and the stock control of the company’s vasttea blending plant. This was highly time-criticalwork, and the applications team had becomeincreasingly concerned about time spent onadding features while the workload expanded.

In hindsight, the extra work, accepted at atime when there was no backup, amountedpotentially to at least as great a blunder as themagnetic tape fixation that had cost two years.But the LEO system—inherently frail due to itsseveral thousand vacuum tubes—met its obli-gations, although there were occasional crises.When there were faults, day or night, the main-tenance engineers generally rapidly broughtthe system up again, sometimes with program-mer support. It became a LEO Computers tra-dition that we could not say that it was thecomputer’s fault if there were a delay. The com-puter’s reputation must not be tainted.

April–June 2003 9

Figure 5. David Caminer at a LEO II console, late 1950s. (Photo from DavidCaminer’s personal collection.)

Success: LEO IIIThe trip to the US that Pinkerton and I took

in 195812 stimulated further system develop-ment. Rightly or wrongly, it seemed that wewere still a long way ahead on systems archi-tecture and business applications but werefalling behind on technology. As a conse-quence, one of the first steps on our return wasfor Pinkerton to plan a magnetic core memoryin LEO II in place of the interleaved delay linememory of which we had been so proud. Itincreased speed but, more importantly, itreduced the space requirements. Most impor-tantly, it gave experience in handling a solidstate memory in preparation for the LEO III,which appeared a year after the last LEO IIdelivery to Ford.

Much has been written about LEO III, andmany in the LEO camp believed it to be threeyears ahead of the corresponding IBM system infacilities and performance. The system parame-ters are set out in Peter Bird’s book.1 The newsystem fully met the “second generation”description. It was designed in diode and tran-sistor technology for the logical circuits, mag-netic core storage, and transistor registers. Thewords were longer, and the fast memory had acapacity more than 30 times larger than on LEOI, giving a potential size (massive for those days)of 65,000 words. Add time was 40 times faster at34 microseconds. The I/O system was now thor-oughly integrated into the mainframe instead ofbeing latch-ons at the front end and back.

In addition to these speed and capacityincreases, which reflected the continuedadvance of solid-state technology, there weretwo major innovations. The first was the use ofmicroprogramming to build the instruction set.This was a technique derived from Wilkes, whorecounts how he thought out the idea on theway back from a visit to the Whirlwind com-puter at the Massachusetts Institute ofTechnology.13 The technique involved a wired-in array of basic actions, which could beinvoked by microprograms held in read-onlycore matrices to form the instructions availableto the programmer. An incidental outcome wasa further speed improvement as references tothe memory to pick up instructions were sig-nificantly reduced.

Pinkerton quickly latched on to the newtechnique, and John Gosden (from the consul-tancy team who worked on the system softwareaspects of LEO III) joined with the engineers indetermining what the instruction set shouldbe. Gosden brought with him the experiencegained by the whole team working on the fullspectrum of business applications as well as a

remarkable variety of service work. A rich reper-toire of 92 instructions was developed.1 Eventhis large number could be added to withouttoo much difficulty if, in a particular program,a macro sequence was frequently employed.

The second innovation was multiprogram-ming, or time-sharing. This enabled several sep-arate programs to run concurrently with nodanger of interfering with each other. The cal-culator speed had advanced more than that ofthe peripherals, and this feature enabled it tobe fully exploited.

To complement the hardware, Gosden hadsketched a software package aimed at enablingthe programmer to concentrate on what thecomputer was asked to do rather than to con-sider how the system would do it. By Microsoft’sWindows standard, the operating system wastiny but efficient, occupying only a relativelysmall part of the memory. Because it dealt onlywith essentials, it was free from the faults thatso often plague modern operating systems.

The integrated hardware and software LEO IIIsystem was backed by a remarkable cadre of con-sultants. Several of these consultants had beenwon from further talent trawls through Lyons’existing personnel, but the majority came freshfrom university. Oxford and Cambridge hadbeen particularly cooperative in steering someof the more venturesome students in LEO’sdirection, and the team reflected this quality.None of the recruits brought with them anycomputer knowledge. Yet all had become profi-cient programmers and learned to write clear,concise specifications before developing appli-cations of their own. Although they wereresponsible for sales, they received no bonus orcommission. They were called consultants.

Peter Hermon, a senior consultant, wholater designed and implemented the award-winning reservation system for British Airways,declared:

Much of the excitement came from the thrill ofhandling a job in its entirety, definition, storelay-out, coding, testing, operating instructionsand user liaison. Not for us the drudgery of theassembly-line coder. In our enthusiasm (andsometimes naivety), nothing was sacrosanct,nothing impossible.2

DemiseSo, in the early 1960s, LEO Computers saw

itself ahead of other vendors not only in thesuitability of its equipment for large-scale officework, but also in its ability to go into largeinstitutions, analyze their needs, and put theircomputers to do productive work.



Leo Fantl and Peter Hermon have alreadybeen mentioned. Among others in animplementation-tried team were Frank Land,who became chief consultant; Doug Comish, awartime Cambridge footballer; John Aris, lateron the Director of the National ComputerCentre; and Ninian Eadie and Mike Jackson,both international-class sailors.

Why, then, with a lead in systems designand such strength in consultancy personnel,did LEO Computers disappear from the marketwithin a few years? In hindsight, there weremany reasons. The home marketplace was toosmall. British industry and administration weretraditionally less venturesome than theirAmerican counterparts. The British govern-ment was resolutely noninterventionist andtimid in its own computerization. LEOComputers, unlike its main competitors suchas IBM and International Computers Ltd. (ICL),had no sizable market base on which to capi-talize. And, perhaps equally important, theLEO team was so convinced that it had theright answers that it tended to lose sight of thefact that the ordinary businessman had notaste for adventure and risk in a field that, atthe time, was not essential for his operations.

Further, the LEO management team, socohesive in the past, was now divided—differ-ences were not aired and they were notresolved. At the operational level, Pinkertonand I were convinced that we had produced astate-of-the-art system, both hardware and soft-ware, for which there was a significant market.In our coast-to-coast tour of leading US instal-lations,12 we had been startled by the numberof large and expensive IBM, Univac, and othersystems that were either employed or were ontheir way. We were unable in many cases toidentify how the systems had been justifiedeconomically, but we learned to appreciate thatit had already come to be regarded as axiomat-ic that every large business had to have its suiteof large computers. The computer had becomeas much fashion as utility, and it stood to rea-son to believe that this fashion, like so manyothers, would cross the Atlantic.

However, at the other end of the manage-ment spectrum, Lyons’ Anthony Salmon wasbecoming concerned that the growing demandsof LEO Computers were threatening to becomegreater than the parent company could bear at atime when the core catering activities werethemselves going through a challenging post-war period. With the need for continuousinvestment both in development and in workin progress, the subsidiary had become burden-some. LEO Computers’ future was becoming a

strategic business matter to be resolved in theLyons’ manner at the highest level.

Simmons’ attitude also played a role. Sim-mons had warmly supported the building ofthe first LEO because he saw it as the answer toLyons’ own compelling office needs. Similarly,he had supported the sale of copies of the LEOII to chosen outsiders because it was a way ofspreading the development costs, rather thanbecause of any inclination to go into the officemachine business. Now, he welcomed LEO IIIbecause it met Lyons’ ultimate ambition forintegrating all the applications for all its divi-sions in what was described as its master plan.11

However, backed by his experience as the headof the Office Management Association, the rep-resentative body of UK office administrators,Simmons was unconvinced that there was anysubstantial number of users capable of con-structively and creatively using large-scale com-puter systems.

Another line of thought at this time hadbeen to concentrate on the service bureauactivities. The successful relationship withFord, Kodak, and many others had indicatedthat there was a large market to be tapped ifusers could be weaned from their distaste ofbeing experted by what was still widely seen asa tea shop company. The case for this approachhad been strengthened by a partnershipformed with a leading South Africa gold-minesorganization. After a world survey, Rand Mineshad decided that LEO III was what they want-ed, but, in view of the geographical distance,proposed that it and LEO Computers shouldjointly offer a bureau service. Thompson wasparticularly involved with this proposition, andhe agreed to this.

So, with these cross-currents, Salmon set outquietly on the threshold of the LEO III launchto seek ways of reducing Lyons’ exposure to thecomputer trade. In discussion with Lazard’s, thecompany’s merchant bankers, it was discoveredthat English Electric, one of the UK’s leadingelectrical engineering companies, had reachedthe same conclusion. Both Lyons and EnglishElectric thought that additional partners wereneeded, so Salmon and Gordon Radley, anEnglish Electric director, arranged a tour ofEurope to see whether these could be foundamong the several French, German, Dutch, andItalian companies working on different aspectsof data processing. They were unsuccessful inthis quest but formed a good-enough relation-ship to agree to pool their computer sub-sidiaries and share the burden between the two.This, of course, was carried out in secrecy toavoid the danger of users and potential users

April–June 2003 11

learning that the future was uncertain. Themerged entity was named English Electric LEOComputers.

It was an unbalanced partnership in thenegotiations that followed. Radley was a dis-tinguished engineer while Salmon made noclaim to be knowledgeable about computers.Despite his affection for the subsidiary forwhich he was responsible, it was, in the end,just another business. This did not put Salmonin an ideal position to defend LEO’s interests.All the discussions seem to have been aboutfinancial arrangements and appointments,with nothing about the choices that the newcompany would have to make from the port-folio of computers that it had now inherited.The chairman and managing director were tobe from English Electric, and the electrical engi-neering company was to be given the lead indeciding what systems to market.

The merger with English Electric wasannounced in February 1963, nine monthsafter the LEO III installation in London andeight months after the LEO III installation inJohannesburg. After a short period, LEOComputers sold its share of the merged com-pany to English Electric and recouped itsexpenditure over the years. LEO was droppedfrom the company’s name.

Thompson, who seems to have been onlypartly informed of how the negotiations pro-gressed, now found himself limited in terms ofdeciding where to devote his energies. Afteryears of reporting to Simmons, he could notbring himself to report to the newly appointedmanaging director Wilf Scott, and after a miser-able three years, he left the company to join oneof the largest LEO user companies as an adviser.

Meanwhile, unaffected yet by the merger,LEO computer development and marketingwas reaching new heights. A top-of-the-linemachine, the LEO 326, was unveiled in 1963.The LEO 326 had a memory cycle time andoverall speed five times faster than the basicLEO lll. This closely met the needs of the UKPost Office, which ordered a number of themto build a service chain all over the country. In1964 they placed a further order for five 326s.This was the largest order ever placed in Europeand large by any standard. Paradoxically, thisorder came a month after Lyons had sold backto English Electric its share of the joint compa-ny. In this deal, all Lyons’ development expens-es were recovered. It was a sad day, though, forthe engineering and applications teams thathad labored so devotedly over the years.

In 1965 the LEO high point was reached—17 systems, many of them top of the line, were

delivered. In the following years, 14 more weredelivered despite the severe dislocation causedby the merger and disposal of Lyons’ interests.But the ensuing downturn was inevitable;moreover the pool of consultants was runningdry. Several had no taste for what had becomethe norm—merely responding to invitations totender with little opportunity for any in-depthstudy of the customer’s problem—and leftEnglish Electric to work for large users.

There was yet another blow. For all itsstrength as an engineering company, EnglishElectric soon decided that it could not providethe resources necessary to design a replacementproduct line. Accordingly, English Electric decid-ed to jump the gun and import a design fromRCA with which it had cross-licensing arrange-ments. To build the larger units in its Spectrarange, RCA’s design featured advanced multi-layer circuitry. To meet the needs of LEO cus-tomers, however, a much larger top-of-the-linemodel was designed by the UK engineers, andthe operating system was entirely redesigned.

Apart from anything else, English Electric feltthat adopting the RCA model would bridge thegap between the English Electric and LEO wingsof the organization. It was an economical wayout, although this meant that the LEO brandand the English Electric KDF9, a good scientificsystem, had to be abandoned before their time.

It was a tough end to the LEO story. The lastfew years were a contradictory maze of increas-ing performance, higher deliveries, and suc-cessful implementations interlaced with thedead hand of a merger. All that remained wasthe LEO message of creativity and integrity car-ried by LEO people wherever they went, allover the world.

References and notes1. P. Bird, LEO: The First Business Computer, Hasler

Publishing, 1994.2. D.T. Caminer et al., LEO—The Incredible Story of

the World’s First Business Computer, McGraw-Hill,1998. (This is the revised US edition of Caminer,User-Driven Innovation, McGraw-Hill, 1996.)

3. G. Ferry, A Computer Called LEO, Fourth Estate,2003.

4. J. Aris, “Inventing Systems Engineering,” IEEEAnnals of the History of Computing, vol. 22, no.3,July–Sept. 2000, pp. 4-15.

5. F. Land, “The First Business Computer: A CaseStudy in User-Driven Innovation,” IEEE Annals ofthe History of Computing, vol. 22, no.3, July-Sept.2000, pp. 16-26.

6. T.R.Thompson, The LEO Chronicle, record ofevents concerning LEO compiled by 1947–1962,D.T. Caminer personal papers.



7. P. Bird, The First Food Empire, Phillimore, 2000. 8. J. Pinkerton, Computer Bull., 1986.9. J.A.N. Lee, “Derek Charles Hemy” (obituary in

Biographies dept.), IEEE Annals of the History ofComputing, vol. 23, no. 3, July–Sept. 2001, p. 85.

10. D.T. Caminer, personal papers and applicationspecifications.

11. J.R.M. Simmons, LEO and the Managers,MacDonald, 1962, p. 19.

12. J. Pinkerton and D.T. Caminer, Diary of Visit toAmerica, 1958, D.T. Caminer personal papers.

13. M. Wilkes, Memoirs of a Computer Pioneer, MITPress, 1985, p. 143.

David Tresman Caminerjoined J. Lyons & Co as a man-agement trainee in 1936. Afterwar service from 1940 to 1943,he became manager of Lyons’systems research, moving to theLEO project in 1950. He becamea director of LEO Computers in

1959 as manager with responsibility for marketingand consultancy. Later, at ICL, positions includeddirector of systems software requirements for the newrange and director of new range introduction. Hecompleted his formal career as project director for theEuropean Economic Community’s implementationof a computer and communications network. He wasdecorated on retirement in 1980 with the O.B.E.[Order of the British Empire] “for services to BritishCommercial Interests in Luxembourg.”

Readers may contact David Caminer at [email protected].

For further information on this or any other com-puting topic, please visit our Digital Library athttp://computer.org/publications/dlib.

Editor’s NoteAdditionally, there are two archive sources:• LEO Archive, National Archive of Historic

Computing, University of Manchester,http://is.lse.ac.uk/leo/.

• John Simmons papers, Modern RecordsCentre, University of Warwick, UK.

NewfromTheMITPress

now in paperback

Unlockingthe ClubhouseWomen in ComputingJane Margolis and Allan Fisher“It can’t be easy to write up a pieceof rigorous, academic research intosomething practical and readable,but the authors of this book havedone just that.” — First Monday182 pp., 5 illus. $12.95 paper

To order call 800-405-1619.Prices subject to change without notice.

http://mitpress.mit.edu

FromAirlineReservationsto Sonic the HedgehogA History of the Software IndustryMartin Campbell-Kelly“Should be the starting point foranyone interested in the history ofthe software business.”— Steven Usselman, GeorgiaInstitute of TechnologyHistory of Computing series392 pp., 23 illus. $29.95

A Historyof ModernComputingsecond editionPaul E. CeruzziFrom the first digital computerto the Web—a story of individuals,institutions, and the forces thatled to a series of dramatic transfor-mations.History of Computing series488 pp., 69 illus. $22.95 paper

behind the curtain at leo: a personal reminiscence

Documents