A DVD of Alan King's Bletchley Park lectures titled 'LEO: The First Business Computer' on 20 January 2005.

King discusses the administrative efficiency of J. Lyons & Co., the early days of LEO I, the programs and jobs that were carried out, the progression of LEO Computers Ltd, and its subsequent end. He also discusses his role in the company and the demise of J. Lyons & Co.

Transcript :

[HOST] Well, good evening, ladies and gentlemen. Nice to see you back. Welcome to the first lecture of
2005. I hope we shall see you again for many more. The subject matter tonight is LEO,
the first business computer, and our speaker is Mr. Alan King. Now, he knows much more about
himself than I do, so he's going to tell you a little bit about his background and what have you
before we start. So I'm sure he'll give us the most entertaining evening, and it gives me great
pleasure to introduce Mr. King.
[AK] Thank you.
Ladies and gentlemen, good evening. I feel privileged to be here, particularly in this lovely room.
Thank you for inviting me, or thank you for coming.
I have a slight dilemma because I could talk for hours and hours and hours about Leo,
the first business computer. Is there, by the way, anybody here who shares my background before I
say too much? No, nobody who worked on the LEO project? Oh, good.
Because I feared when I was invited that you would expect me to talk knowledgeably about valves and
wiring and electronics and other things. And although I will endeavour to say something about the
structure of the machine, I'm going to cover other topics as well.
So effectively, to give you a script, a brief prologue about me,
then I will talk about how many of you used to go to a corner house or a tea shop when you were
younger.
So you're probably wondering, if you don't already know, how a leading caterer came to build the
world's first business computer. Well, that's the first topic of four that I shall try and give you
some background to. Then I'll say a little bit about the design of the machine, how it was put
together and how it worked. And I thought perhaps we might cover... Some of the challenges that the
team of people who worked on it faced, because it was the first. There were no precedents.
They made it all up as they went along. And I thought perhaps I'd tell you something about what Joe
Lyons used the machine for, because it wasn't to save money and get rid of clerks.
And then there'll be an epilogue.
Right, well, if that's all right with anybody, and if it's not, sorry, too late.
How come it was J. Lyons. Well, Lyons was started in the late 1890s by a firm called Salmon and Gluckstein,
who were leading tobacconists. Now, I'm sure that although lots of you, like me,
are a bit thin on top, none of you remember Salmon and Gluckstein tobacco shops in the local High
Street. But they were a very leading tobacconist, who ultimately became part of Imperial Tobacco,
and that was formed in the... early part of the last century, to compete with the Americans.
And they wanted to diversify, so they decided to diversify into catering, initially for
exhibitions, and because they wanted to hide the fact that it was Salmon and Gluckstein who'd set up
this new business, they persuaded a distant cousin, several times removed, called Joseph Lyons,
to head up the organisation. So he did. And they were very successful at catering,
did large exhibitions in Newcastle and London. And I think the last time I looked in the Guinness
Book of Records, which was some time ago, they still held the record for the largest sit-down
meal, which was umpteen thousand people in Olympia, somewhere in the 1920s.
And they then moved from outdoor catering into tea shops and hotels. And then they started
manufacturing food to supply their own tea shops and hotels,
but also a local retail trade in West London, which rapidly became nationwide.
And they did tea and coffee and bread and cakes and ice cream and frozen food,
and the list goes on and on and on. And as was the custom in the 20s and 30s,
business philosophy was vertical integration. So because they had a lot of menus and other things
printed by the Hogarth Press, they took them over. Because they had a lot of linen that was
laundered by Hayes Laundry, they took them over. Because they had lots of vehicles that were
provided in the service by the Norman Car Group, they took them over. And again,
there's quite a long list of organisations that most people don't realise were at one stage part of
the J. Lyons empire. And then they got into computer manufacturing. Why did they get into
computer? Because you can see the logic of laundries and other things, but not necessarily
computers. Well, it was because the business was always high volume,
low profit margin. I won't bore you with a lot of figures. But if you went into a tea shop and had
a cup of tea and a bun, they got a farthing profit on that, one quarter of a D.
So it was really very important that the business was under tight control. And that became
increasingly difficult as it got bigger. Anybody like to guess how many tea shops there were at the
most?
A few more at the most, yes. I think 254 was the most. Very good.
You printed it all off, did you? I think 254 was the maximum number.
So in the early 1920s, they recruited a Cambridge graduate called John Simmons.
And he headed up a thing that became known as the Comptroller's Department,
which provided clerical services to most of the... manufacturing,
distribution, selling operations.
And then in 1931, he set up within that controller's department a thing called the Systems Research
Department, which looked into clerical administrative efficiency.
So Joe Lyons became a leader in a number of things you wouldn't have expected. Clerical work study,
organisation and methods, operations research. I think they were also the first organisation in the
world to use microfilming for business records. But certainly they had a microfilm project going in
the mid-1930s.
They had a concept of regular departmental audits, they did,
and to ensure optimum efficiency of information going to management.
It was always part of the philosophy that you built in feedback to the clinical operations,
so you told people how well or badly they were doing, and that you provided them with information.
not figures, information, not data.
And then in 1947, just after the Second World War,
two of the very senior people called Thompson and Standingford from the Systems Research Department
went to the States. And they went to look at what our American cousins across the water were doing,
other than getting enigmas out of submarines.
They went to see what they were doing with office equipment and administrative systems.
And they were shown lots of bits and pieces, including the Harvard Mark I computer.
And they had the insight to say, this machine,
if it could do clerical work, will enable us to bypass punch card. Anybody work with punch cards?
Equipment, dreadful things. I feel sorry for you. So Standingford and Thompson said,
look, if we could get one of these things to take the place of the central tabulating unit of a
punch card installation, it's going to give us a lot more power and flexibility. So they came back
to the United Kingdom, and the Americans had given them, said, why don't you go and talk to a guy
called Morris Wilkes at Cambridge University? So they did. And he was playing around with a stored
program computer called EDSAC.
And so Thompson and Standingford went back, reported to Simmons and said,
we really think we ought to go into computers to tackle our administrative workload.
And they looked at a number of options, like buying one from America and decided they couldn't
afford it, or trying to persuade Cambridge University to use theirs. And eventually they said to
the Joe Lyons board, and this was in... 20th of October 1947.
If we can show that this thing at Cambridge works, will you let us build one? Will you pay us to
build one of our own? And the J-Lion board, who were a very imaginative bunch of people,
said, yes, OK, fine. You can have £3,000 and one of our engineers to help Cambridge out.
And in return, if they can show that it will work, reasonably, on a stored program,
then we'll pay to build one.
So this is what happened. They put £3,000, which is about,
I would think, somewhere around £60,000 in current money. They put it into EDSAC, no guarantee of
any return at all, and estimated that...
build a commercially oriented version, if EDSAC did work, would require about £100,000 but would
generate £50,000 a year of extra profit contribution. That was a massive sum to commit in the late
1950s.
But that was their decision.
And on the 6th of May 1949, EDSAC worked. It did.
It carried out a stored programme. There was immediate phone call to J. Lyon's head office at
Cadby Hall in West London, and a note was passed into the boardroom, because the board were meeting
at the time, saying, yes, yes, yes, it works. And the note came out with a signal, with a
handwritten message on from the chairman, which said, you have authority to go ahead. So they did,
and they built one. They put together a team of people under John Simmons with Thompson overall
command. A gentleman called David Caminer, who was one of the top guys in the systems research
department I've already mentioned. He was put in charge.
A guy called Derek Hemme was made the first programmer. And they recruited a man called John
Pinkerton, another Cambridge graduate, to help them build it.
And they
suggested three applications, sales invoicing and inventory management, letter writing,
which we would call word processing now, and payroll. And they started to build the machine,
and it was actually demonstrated in February 1951 to Princess Elizabeth, who is now our queen.
And some systems design and payroll development proceeded in parallel, and on the 28th of November
1951, they put a business application live on it. It was chosen because it was low input and
output. It was not critical to the ongoing operation of the daily business.
And it involved taking the production out of the bakery plants, breaking it down into raw and
wrapping materials, working out what materials ought to have been used for the amount of production
and doing an evaluation of how well the production process has functioned in terms of what it
ought to have taken to produce this much Swiss roll and that much cake and that much bread and how
much it had actually taken. And that went live on the 29th of November 1951 and is quite
indisputably the first business use of a computer anywhere.
Whether you think that's been good for humanity or not, 50 years later we won't go into it.
Practical usage on other things was delayed because during 1952-53 they were struggling to find
the right devices to handle high volume information going into the machine and high volume results
coming out. And they experimented with magnetic tape decks. I won't embarrass anybody who might be
here by telling you which organisation they worked with. But effectively it was really a failure.
So at the last minute they had to go back to using punch cards and paper tape to get the stuff into
the machine and punch cards and good old-fashioned tabulating machine printers to get the results
out.
And it worked in a configuration that they thought could handle bulk business applications on
Christmas Eve 1953.
And the first major application, which was factory staff payroll, then went live in February 1954.
And to put that in context, that was the year that food rationing ended.
It was the year that Bannister ran the four-minute mile. It was the year that the French army was
defeated at Dien Bien Phu. It was the year that the USA exploded the second hydrogen bomb at Bikini
Atoll. It was the year that British Petroleum was formed from the Anglo-Iranian Oil Company.
I have to make clear to you a confession that I didn't join until 1958,
by which time it had been running for four years.
It was an indication of the people. I remember when I joined,
of course there were no computer degrees. I don't think there was a British Computer Society,
I can't quite remember.
My colleagues, there were a couple of lawyers who tried to be barristers and hadn't made it,
hadn't been able to build up a clientele. There was a guy with an agricultural degree from Reading,
a couple of mathematicians.
I'd just finished a classics degree at Oxford. If any of you know the peculiarities of our old
places, I read greats if that means anything to you.
And I'd been expecting to do national service and then I found they were phasing it out so all of a
sudden there I was coming up to finals and no job to go to. So I wrote to Joe Lyons because they
were advertising for trainee food buyers and I thought that sounds quite interesting.
And I wrote back and said, dear Mr King, we filled all those vacancies, but we've got openings in
systems research for computer programmers. And I remember I wrote back and said, I'm sorry,
but I've only got O-level maths and no physics. I don't really think it's the same. And then I had
a letter saying, dear Mr King, you obviously don't know the first thing about computers. You'd
better come for an interview.
So I went for an interview, did an aptitude test. Had an hour's interview with the personnel
director where we argued furiously about Freudian psychology, which he thought was marvellous and I
thought was rubbish. At the end of which he said, well, yes, I think we'll take a chance on you.
Can you start next month at £700 a year? So thinking that I wouldn't dare go home to my father and
then fiancé and say that I still hadn't found a job, I said, done.
So there it was. So I wasn't the first LEO programmer, LEO-1 programmer, but because when the
machine three or four, the last three or four years of LEO-1's existence, I was responsible for
the maintenance of the J. Lyon systems that were still running on it. So I am quite indisputably
the last LEO-1 programmer.
I wondered at the time of the millennium bug whether I would be able to make lots and lots of
money. But unfortunately there were no LEO One programmes around anymore. I'll speak a little bit
about what the machine looked like. Oh, by the way, I should have shown you this.
If you want to know more than I have time to tell you this evening, there are a number of further
reading things there. I've got four of the books here. Do please have a look at them afterwards.
I do trust you not to take them away because they're mine.
If you want one which is not the best read, but certainly gives you the most information,
including something about the electronics, then this one by a guy called Peter Bird would be it.
If you want one written about what was involved and how they did things and how they thought about
it, this one by David Caminer, who I've mentioned, and some colleagues.
If you want one that's the best read, because it's written by a good wordsmith and it's a good
read, and I know somebody had it, then this more recent book by a lady called Georgina Ferry is
very good. Somebody brought one with them. Yes, that's right. You agree, it's a good read.
And this, which is out of print, I shall mention later on this evening, is a book by the man I
mentioned called John Simmons.
Pinkerton and his team...
decided that they'd base their commercial computer on the EDSAC design. And EDSAC,
by the way, stood for Electronic Delay Storage Automatic Calculator. It did.
And the storage bit is important. You know how it's called LEO, by the way?
Lyon's Electronic Office. And John Simmons always used to insist that it wasn't a pun on LEO,
the Latin word for a lion. Nobody ever believed him.
But it's a lot more imaginative than EDSAC. So the guts of the machine in many ways was its
storage. And it was... I told you I didn't have any physics qualifications.
Of course, if you're going to keep a train of pulses in order to look at it and get at it,
you'd need a bit of wire probably three times around the universe.
So during the Second World War... Designed to clean up radar signals,
a guy called Tommy Gold, who later became famous with Hoyle and Bondi for the steady state theory
of the universe, he came up with the idea of sticking a tube of mercury into the circuit,
the piezoelectric thing at each end, and converting the electronic pulses into sound waves
travelling down the mercury. and that delayed the circuit time of the pole strain sufficiently long
that it became a viable proposition as a form of storage and the tubes were actually five foot four
inches long with a 0.001 of an inch tolerance they had a one inch internal dynamometer and half a
ton of mercury
There were 576 pulses in each tube or unit, one microsecond apart,
and you could access that as 32 compartments or separate numbers,
each storing 18 bits or 16 double numbers, each of 32 bits.
EDSAC had 32 of these tubes, but LEO had 64, and that provided you with 2,048 numbers.
So that was the storage of the machine, so you begin to get an idea of what Caminer and his
colleagues were facing in terms of design problems trying to put a payroll into a 2K memory.
Bet you can't get windows into a 2K memory.
Bit 18 was used as a separation bit because the electronics were a little bit dicey.
And if you were trying to plug into an 18-bit stream in the 500-odd going round the wire and the
mercury, you could get a slight variation. So the 18th bit was a sort of extra,
just a separation bit. So each of these compartments could hold...
2 to the 16 minus 1 which is 65535 as a binary number and also of course they held an action an
instruction to the machine so there were 32 actions which could address the 64 tubes and 32
compartments within each tube the
access time was 500 milliseconds give or take a few depending upon quite where the pulse strain was
at the gate at the end of the delay line. Numeric only of course,
no numbers at all.
And then the machine had 200 logic units.
If I can find them quickly. I've got one or two photographs.
I'm sorry these are not of the best, but it was a long time ago.
Can you all see that all right? The bottom are the mercury delay tubes.
And they were actually under the floor. The computers, they were actually on a raised floor and
they were under the floor and they were popularly for that reason known as the coffins that they
were in.
And then you had, yes that's a good one to show you,
there were 200 of these racks and it was all built with valves and wires,
no microelectronics or anything. I'm sorry I can't, as I've already apologised to you this evening,
tell you too much about the electronics. But there were 200 of these racks. To reduce the risk of
failure, all the racks were removable and you could take one out when it failed and plug one in.
You didn't actually have to mess around with your soldering iron and fit a new valve. You just took
the rack out. And down here is the false floor where the coffins with the mercury delay line
storage were.
The whole thing generated, the whole thing used, hang on if I check my notes, 30 kilowatts of
power.
So you had air conditioning ducts, which you can see up the top.
And standby diesel generators.
For those of you who are interested in these things compared with an Intel 4 processor chip,
simple arithmetic and sequence change instructions took 1.3 to 1.6 microseconds to work and something a
bit more complicated like multiply took 6.7 microseconds that seemed bloody sorry that seemed very
fast in the late 50s of course when computers work in nanoseconds now it doesn't seem quite so fast
but nonetheless it was um it was very impressive it had a control console which looked like that.
It did. One or two interesting things. These were oscilloscopes that engineering controls.
These were oscilloscopes that let you look at any of the compartments within the tubes and the
digital pattern that was in it. And these buttons here, and this was very naughty,
as programmers we had our wrists slapped if we were caught doing it. But if you were running a test
and it went wrong... and you could locate where it was. You could call up the wrong,
you could call up the pattern of the order as an 18-bit pattern on these oscilloscopes.
You could then, here, you could then button in the correct 18-bit order,
and then you could then press one of these buttons, which was called stacking store,
so you could actually... put the correct sequence of instructions into the machine store in flight
as it were while you were running a test program greatly frowned upon but if the boss wasn't
looking one was inclined to do it and that there was an emergency stop button and there's one
interesting story where you can guess what's happening the cleaning lady with her broom one day was
doing this and the broom handle hit the emergency stop button And, of course, then those of you who
are engineers will appreciate you were not at all popular if you hit the emergency stop button when
you didn't have to because it didn't do nice things to the valves at all to have the power cut off
in a microsecond or so.
And then, as I've already said, the input and output devices, they had to go for punch cards
because the magnetic tape wouldn't work reliably.
Let me show you that one, which is off a mouse mat. It's been taken as a mouse mat.
There you can see the racks of logic units with their covers on and the control desk up the end and
the false floor where the coffins were underneath.
This is a very poor quality photograph where you are looking across here at the racks and there's
the main control console. And this was a subsidiary control console here.
for the operators and these were the punch card readers that were used to feed information in and
the printer that was used to print information out and the paper tape readers that were used to put
information in. It had three channels for input because Caminer and his people decided that was
enough to run a clerical operation. One channel for brought forward record like an employee file or
a stock file and a couple for current information like hours worked or stock issued from the from
the stores and then it and you so you had one channel with a 200 card per minute 80 column card
reader on it one channel with a franti five hole paper tape i think it was 1500 120 characters per
second tape reader And a third channel that was switchable, so you could have two card inputs and a
tape, or two tape and a card going at the same time. The output was a punch card punch,
80 column card punch going at 100 cards per minute, or a line printer,
80 column line printer, which only had numbers on.
And that went at 100 lines a minute, if you were lucky.
So that's all I was proposing to say this evening about the machine.
And I thought then we might have a look at some of the design challenges that Pinkerton and Caminer
and their people had to face. I've already said there were no existing administrative applications.
Nobody else in the world was doing this in 1951. So every aspect of using a computer for business
purposes, they had to invent.
And they relied greatly upon the Systems Research Unit, which had built up,
you may remember, this excellent reputation for using O&M clerical work-study techniques to
ensure that clerical operations functioned, administrative operations functioned as efficiently and
rapidly as possible and reliably as possible. So they built up a whole hierarchy that I had to
learn when I started, whereby the first thing you did if you were computerising an application was
to go away and find out how it worked and talk to the managers and talk to the people and you
charted it all and you agreed with them what happened at each stage in the clerical operation what
the inputs were what they did with it what information then passed on to the next desk and so on
and so forth and then you designed a set of programs to mirror this with flowcharts and only then
did you sit down and pick up your coding pen and write Computer code.
I'll say about that in a minute. Then one of your colleagues desk checked it all.
In other words, you might write a routine to do payroll or stock control and then you'd pass it to
a colleague who would pretend to be a computer and obey it.
And only then, because computer time was so precious, did you actually prepare punch cards and
paper tape of test information, trial information, which tried
to put through the machine every combination of data that it was likely to meet in operation and
that's why people like me are so nervous when you get into an aircraft and you think that it's
flying under program control because you will realize of course instantly that
a program, as I'm sure you all know, is like a column of instructions, that the machine being a
dumb idiot just works its way down the column of instructions. And if it comes to a point of time
where it says, well, I want to test something, is this a 0 or a 1?
At that point of time, there's a test and there are then two alternative paths through the logic.
If you get a second test, you get four alternative passes. Most of you will quickly be with me that
if you have 20 points of test, there are over a million paths through that logic. There are.
So testing them is a bit difficult.
But one did one's best. And only when you try to be certain that you've got all the bugs out were
you allowed near LEO I to try your program routine and see if it worked.
So actually running it on the machine was really... to check that the errors were all out rather
than to find the errors and later on in one's career people like me felt a great release when all
of a sudden you were provided with a machine that worked that much faster and had a megabyte memory
rather than a 2k memory because then you could afford with the machines running that much faster,
multitasking, running half a dozen programs simultaneously. You could afford to debug your program
live on the computer. But you still weren't allowed to use the stack install buttons if the boss
was looking.
There were space constraints. You try fitting a payroll routine, a calculation of gross pay,
net pay and all the rest of it into a 2K memory with all the data of the employee record and that,
and it doesn't go. So nearly all applications had to be divided into a number of linked programs
that were run in sequence with intermediate results in the form of punch card decks passing between
them with perhaps a punch card sorter putting them into a different sequence and i've actually here
can
i borrow my wife at this stage please i've actually here and it's very flimsy got the
You go that way, I'll go this way. Now this is the flowchart of programmes for the payroll routine
that was the first real thing that went live in February 1954.
So it says tax year end, amendments where you altered putting starters and leavers and wage
increases, gross and net pay calculation,
club statistics, reconciliation, starters and leavers schedule and so on. So the actual payroll
involved several hours of running linked programmes with all of these intermediate results files
passing between the programmes. That was the only way that you could actually get a real-life
complicated business operation into a 2K machine.
Thank you.
Caminer and his people had to think of all sorts of things. You've got a 2K memory and you've got
this concept of a program being a column of instructions that you obey starting from the stop.
Well, if you want to add up 20 figures and find a total, then saying pick up the first one and put
it into an accumulator, add the second one in, add the third one in, takes 20 instructions. It's
much quicker if you can have a little loop and you can modify the address of which one you're
picking up. So you have an instruction that says add the contents of a compartment into an
accumulator from a particular address. And the next instruction then says pick up one,
add it to the instruction address and stack that new instruction back.
Sorry, I'm explaining this badly. You develop this concept
like laying the pavement you're walking on as you're walking along it so do most of you i mean so
if you had a common routine that appeared for half a dozen times in the program that would be put
into what was called a subroutine so then you had to sequence change each point where you wanted to
use that subroutine of common instructions you had to leave the column of commands change sequence
to the subroutine, obey the subroutine and then go back to the right place.
So what you did is, when you wanted to enter the subroutine, you picked up the address of the
storage compartment where that instruction was, added one to it, stacked it as the last instruction
of the subroutine, then sequence change to the start of the subroutine, went through the subroutine
and at the end it would automatically go back to... the next instruction after the one where it had
left i think i've got that right right but all of this all of this they had to invent and i take my
hat off to them because they weren't learning from microsoft how to do it they were doing it for
the first time so
um yeah
there were all sorts of other things um there
were 2048 addresses well if you were writing a program referring to the addresses.
I want to go to this compartment to pick up the wage rate, that compartment to pick up an employee
number. I want to sequence change from this compartment to that compartment. compartment you could
write down the absolute store address where you wanted to go but of course that meant every time
you made some change you had to go through the program and change all the store addresses. So they
invented the concept of relative addressing whereby when they were actually writing the coding they
would write reference if there was a set of consecutive compartments that were going to be holding
the employee record, gross pay rate and this sort of thing,
you would write an address relative to the start of that particular type of information.
And then when you took your punch card deck of program and loaded it into the machine to run that
program, you had two special cards on the front called an initial loading sequence.
You did. That changed all the relative addresses into absolute addresses because you fed in the
absolute address for each of these sections that the other addresses had been relative to.
You fed them in when you loaded the programme and in loading the programme, the initial programme
loading sequence changed all the relative addresses to absolute addresses. So that minimised all
the work you had to do when you make any amendment because otherwise you had to go through the
whole of the programme changing all the absolute addresses which was a long time very error prone.
So that was a nice bright idea. How do you think they got the machine to start a programme? Well
what they did was on that, I've taken it away now, on the control console there
were a couple of buttons, I forget where they were, I think they were over here.
forced into compartment 0, the first one of the 2048, forced an order hardwired into that
compartment, an order that said, read a punch card.
It did. And then having obeyed that instruction that had been forced into the machine,
the machine read a punch card into the first 32 compartments,
stepped on its sequence register, instruction sequence register, by one, and therefore obeyed the
instruction in the first compartment as the second compartment, which by then had been read into
the machine. And so it self-started. And it was called bootstrapping,
because you pulled yourself up by your own boots. Do you, by the way, know why computer errors are
called bugs?
Well, there was a lady in America called... oh it's an early sign of elzheimer's when you forget
names um Grace Hopper there's a lady in America who was I think it's the Harvard mark one and she
was playing around with a card they held programs like we did on card decks and a program that had
run perfectly for many a long year suddenly went wrong and they couldn't understand why, and they
investigated and found that a little beetle had got stuck between two cards and blocked up one of
the holes in the cards. So it had been read as a nought instead of a one, and that was a bug.
It's an apocryphal story. I'd like to think it's true, and I'm always told it's true, but never
mind. It's a nice story, isn't it, as to why computer errors are called bugs.
Time passes.
So all sorts of things. You had to work out how long the sequence of instructions you'd written
took to run at one point so many microseconds for this instruction and milliseconds for that one.
Because if your main loop of what you were doing took longer than the card reader took to read in a
card, then your program running time immediately doubled. So you had to make certain that your main
calculation routines took less computer time than a card read cycle took.
That was quite fun. It was all numbers only, of course.
So when you were printing payslips, you printed on pre-printed stationery.
You all remember great big overgrown toilet rolls of the stuff.
When you got a punch card deck, which held the... details, one card for each employee,
the cards were all pre-printed with what the rows were holding,
gross pay to date, tax to date, pay rate, employee number,
unit number and this sort of thing.
But if you were printing other results, you printed it on plain listing paper. But then,
of course, if that was, say, a schedule of starters and leavers, the poor old clerks who were given
the schedule, got an array of numbers.
So you solve that problem by
having one of these, which was an overlay. So they put the overlay like that on top of the
printout. Do come and have a look at this afterwards. There
you are. So you put this over two consecutive rows of ones and noughts and other funny figures on
plain listing paper, which was like what my wife and I held up, plain listing paper and that would
enable you to see that those figures were gross pay to date if you were a clerk. Of course us
programmers got very clever and one of the tricks, one of the skills I still have but like a LEO I
programmer, being a LEO I programmer it's not much use anymore.
You've got to be able to count in binary sterling. You did. So it's a penny,
tuppence, fourpence, eightpence, one and four, two and eight, five and four, ten and eight, one
pound, one and four, two pound, two and eight.
You see? Anybody got, anybody, yes, I know where it can come if I'm looking for a job.
So you learn all sorts of things like this. I rabbit on.
But one thing I will show you, I was going to later on say it was a shame,
but when LEO I was scrapped, it all went. The mercury was sold because it was valuable.
Storage racks were sold because they'd got gold contacts all over the place. It was all broken up.
If you go to the science museum, I think they've got a couple of small buffer delay lines that were
in the card readers and paper tape readers they have. And I think they've got one of the racks,
an adding unit rack. But I remember on yet another office move,
taking piles of system specifications, programming sheets, operator manuals,
user manuals for the clerks, tipping them in black sacks. And they went off to the local Civic Tip
or somewhere. No, they probably went to the site incinerator. They did.
And you just didn't realise that you were throwing away something that has some historical
significance. But I did find some years later that I had kept that. This isn't the original.
I've got that with me if anybody wants to see it. But that is a page of coding.
You'll see it's dated the 27th of August, 1951. And that is a page of coding,
a subroutine that clears item rates. That is a page of coding from the P1 Baker Evaluation
Programme that first ran in November 1951.
Unfortunately, I've shown it to people like David Caminer, who's still alive, whereas Simmons and
many of the other people have passed away. I've shown it to David Caminer, and he can't quite
remember who will have written it. Probably Derek Hemi.
But nobody really knows. But I kept it. I don't quite know why I kept it. But it is certainly,
it was certainly when LEO I closed down in the early 1960s,
it was the oldest piece of operational computer code anywhere in the world.
But that's what it looked like. And these are the relative addresses. 28 was ad, 5 was stacked back
into the store.
Forget what some of the others were. I forget what most of them were. 12 was subtract. 27 was some
sort of test. These were the relative addresses, it was. So this would be the 33rd compartment in
tube 31, this sort of thing. And it was this sequence on paper tape or punch card that went into
the machine in which the initial loading routine then substituted absolute addresses for these
relative addresses.
What did Lyons use it for? Well, they used it to run the business. And they made a promise to their
staff that they kept when it was introduced that there would be no clerical redundancies,
and there weren't. They used it to run a business more efficiently.
So the first thing they did on it was the payroll.
They put 1,700 bakery staff on in parallel in January 1954. You'll remember the machine was
thought to be deemed operational at Christmas Eve 53. So all the programme development had been
going on in parallel while the machine was being got into physically and operational form.
So they put 1,700 bakery staff weekly payroll on in January 54 and it went excellently well.
So on the 9th of February 54 they went live and they stopped doing a clerical operation in backup.
By mid-1954, they'd got 10,000 people having their pay calculated weekly on the machine.
And they stopped there because John Simmons had said, until we have a second backup machine, 10
,000 is the limit. So it had about 30,000 employees at the time. So they had about a third of them
having their pay calculated on the machine by the middle of 54.
But it never failed to produce the pay slips on time. Once or twice I was called out in the
afternoon of Thursday afternoon. It's gone down and I was there all night Thursday with people
breathing fire down my list because a lot of the staff were of Irish descent and there would have
been a nasty riot on Friday. They hadn't got their paper. But no,
I mean seriously it never failed to produce the payslips on time. And it took 1.5 seconds per
employee. And the previous clerical system, which was very, very efficient, because that was all
part of the Simmons Systems Research concept, used to take eight minutes. So that gives you a
comparison.
It did.
There were 250 tea shops, and they'd sold about 150 products. They've all gone now,
so it's no secret that a lot of the food, a lot of the things, used to be cooked at Cadby Hall
overnight and trunked out to the tea shops in the early hours of the morning and crash reheated
there was very little actually cooking took place in the tea shops so one of the very one of the
second system that went onto the machine involved in greatly improving the efficiency of that
operation so what they did is for every 150 items each tea shop manageress had a sort of standing
order that was varied from time to time depending upon summer or winter And then every afternoon
they were telephoned by ladies who sat there with headsets and mouthpieces on so as their hands
were free. And the managers were asked if they wanted to alter their standing order for the
products that would be coming to them in the early hours of the next morning. So the computer
system had all the standing orders on cards. The amendments were fed in on paper tape.
And overnight it calculated the cooking orders for the kitchens at Cadby Hall.
And the stuff all then went out, calculated all the quantities of the food required for each
kitchen, produced all the loading instructions for the van as to what amount of what product was to
go on the van, and it all went out the next morning.
And then the third one, because I'm using up my time. The third one worth mentioning is Lyons used
to make a lot of money on tea. Not on selling tea, but on buying tea profitably on the
international commodity markets.
So they used to have big stocks of what were called original teas and a bit like you probably don't
realize but most tea you buy anywhere now is a blend like a whiskey it's a blend of original teas
made to a recipe and if these recipes if you buy tea badly and you then have to cheat on the recipe
then you know that anything about 13 days later the complaints will go up like that because the
human palate is very delicate and you get lots and lots and lots of gentlemen in Yorkshire who say,
yeah, mother, you didn't get the right tea. And they'd write in and complain and we'd send them a
Dundee cake in a tin. And if you had to cheat on the tea blending one week,
you got... warned them that Dundee cakes would be needed, consumer complaints people,
because you knew almost beyond any challenge that 11 days later the complaints volume would do
that. But the computer was used to keep a record of the tea stocks and to calculate the efficiency
with which the blends were being used in terms of cost, of making that particular batch of Green
Label or Quick Brew or whatever it was that it was being made. So it was that sort of thing.
In view of the time, I'd better move on to my epilogue.
So what happened after 1954?
Well, LEO was used to do a lot of bureau work early on. As early as April 1952,
it was doing calculators for the Ordnance Department. I think it was on Lou...
black black something a missile blue street yes but but we were kept well i wasn't there then in 52
um but the Lyons people were kept out of the machine room and those people who were allowed
in to operate the machine had to sign the official secrets act so there's not a lot of knowledge as
to what it was actually doing. It did weather forecasting, it calculated for British rail
the cost and the best route, least cost route, to move freight from any railway station to any
other railway station. It did the tax tables each year after the Chancellor had sat down by the
next morning to produce the tax tables.
And it did the payroll for Ford at Dagenham. And it did more clerical work and more clerical work
for other people. So in July 54, the Lyons Board authorised a second machine,
which had a somewhat enhanced specification.
And they also set up a company called LEO Computers Limited in November 54 to manufacture and sell
the things. And a lot of staff like Thompson and Pinkerton and Caminer then went and became
directors of LEO Computers Limited. And the J. Lyon system research people took over the
maintenance and development of programmes for the parent company, and that's the department I
joined. I was never in LEO Computers Limited, so you remember we were sharing this outside earlier
on. I always worked for Joe Lyons at Cadby Hall. Eleven LEO IIs were sold from 58 through to 61.
LEO I was finally switched off on the 4th of January 1965.
It was. By then, Lyons had developed a LEO Mark III.
I mean, that was a lovely machine. It had a 16K core store.
It had microprogramming for the instructions instead of being hardwired. It could timeshare,
run four or five operations simultaneously. It had a full operating system,
a cobalt-type high-level language. It did radix arithmetic so it would add up in sterling or
yards, feet and inches or pounds and ounces.
and some 60 of those were sold from 62 through to 69, and the last one went out of operation in
1981.
Much earlier than that, Lyons had decided it had limited capital, and was it a food and drink
company, or was it a computer manufacturer, and he decided it wasn't in computers. Couldn't really
compete with IBM, let alone the Seven Dwarfs. You all know about IBM and the Seven Dwarfs. Do you
know what IBM stands for? We used to have great fun inventing things. No, no,
no, it's better manually.
Or install bigger machines. Or the one I liked most was increasingly big Monopolosaurus.
But we were very rude about IBM.
No, the Lyons really could not compete with the financial strength of people like IBM,
Honeywell and Burroughs. It couldn't compete with their vast customer base of punch card
installations. And with management who didn't like the Lyons philosophy of saying,
let's not do GIGO, garbage in, garbage out. Let's design a proper information system that gives
management information and then we'll computerise it, too many people bought computers because it
was the okay thing to do and then they then relied on their hardware supplier and software packages
to get it to work and we all know about all those systems that don't work properly don't we um so
LEO computers were swallowed up from 63 through 68 in a series of mergers with English Electric,
Marconi and ICT to make what was ICL that's now owned by the Japanese, I think,
isn't it? Fujitsu.
Fujitsu. And the LEO range was abandoned in unsuccessful efforts to match the IBM 3 series and its
successes. J. Lyons continued to do imaginative and leading-edge systems.
I was particularly proud... No, I won't because I've run out of time. We tackled a massive input.
Input was always a problem. How do you get the stuff in?
So we came up with
that.
And what we did was we bought a great big powerful printer called a Xeronic that printed forms and
the information on the forms simultaneously at God knows how many feet a second.
And what we did was we provided, say, a tea salesman with an order form,
which looked like that.
So you printed the name and address, and you sent this to him,
perhaps a week in advance for his following week calls, and you sent him in sequence the people he
had to call on, and you printed the products, tea and coffee,
that he normally bought. You printed the information of the last three calls,
what he'd sold. You printed what the average was. And then over here,
you printed an identity of who it was.
And over here, he then barmarked what quantity they'd ordered at that call. That form then came
back and went through a high-speed barmark reader.
And if there were any queries, you used the same concept. You printed out a query form.
from the DataVet program, and clerical people looked to see what the problem was,
extra line entered or something other, and then they put in what the line number was and fed that
back into a subsidiary input thing, and it was matched up with the original query.
And so you developed a system which we called the turnaround document system.
So the order forms went to the sales reps.
Alders and other document reader data came into LEO and the autolector. Queries went to a query
group. You then did the main daily run and the Xeronic printer printed it out and you went round
and round and round the loop. And that chopped time and time, hours and hours and hours.
But of course, and a lot of money off the cost of getting the stuff into the damn machine.
But of course, equally,
you were helping the guy to sell. Because you were telling him he didn't have to keep any records.
We told him on the order form who he was supposed to call on the next week, the sequence in which
he was supposed to call, and what we'd like him to sell. Oh, and by the way,
it would add in, I forgot to say that, and it would automatically add in things that were on
promotion on top of the original things. So it was a proactive computer system that tried to help
the guys who really generated the profit for the business, the salesmen out in the field.
Am I allowed to be big-headed?
That was my idea. So if I have one thing on my grave that I can mention to my wife at this moment,
it's he invented the turnaround document concept.
So Lyons continue to innovate in computer systems.
if any of you oh for example take the tea blending system we developed a method of getting the tea
tasters to quantify on a numeric scale the taste the hardness the brightness all the key factors
about a tea blend and the tea blending system then be used linear programming techniques to
actually tell the business which original tea stocks to use. So it didn't just keep a record of
what was in stock and how well it had been used, it actually told and then it went yet a stage
further and we put DEC computers as process control computers on the production line and actually
then the whole thing was automatic. If you want to know more about this, and I see there's a
lady... If you want to know more about this, then shortly, somewhere in the other, an academic from
Southampton University called Janet Delve will be speaking about this somewhere in February at the
Science Museum. There's a thing called the Computer Conservation Society, and they have free
monthly, bi-monthly meetings. And there's one somewhere in February,
I think. I'll look it up if anybody really wants to know, where a lady called Dr.
Janet Dell will be talking about the tea blending system. But that, after LEO I,
went on to become, first of all, a linear programming optimization of blending,
and then went on again to interact between the commercial computers in the head office and process
control computers in the factory.
It did. So that's what happened to LEO. That's what happened to Lyon's computer [?].
And then,
towards the end of John Simmons' time, he came up with what was called the master plan.
So the idea was that it would be very similar systems handling depot records,
sales operations control, production control, payroll. and then management accounting for each
operating division. And this is the one for the ice cream division with the information flows
between it.
Sales operation jobs, the ice cream sales operation job is the first one that went up and it was
actually a failure. It had to be taken down because it was supposed to be optimising the stock.
Ice cream's funny.
Summer sales can be six times winter sales. and sales demand can triple overnight if there's a heat
wave. So effectively, you can't possibly have production, distribution and storage, which all has
to be refrigerated, of course. You can't have capacity to meet peak summer demand, otherwise you go
bust in the other 51 weeks of the year, funding the capital involved.
So effectively, you have to make like hell through the winter and guess what the public's going to
want to buy the following summer.
And then you have a problem of trying to optimise the stock in the local distribution depots that
service the local shops. And so that was an operations research type application that tried to
optimise the stock in the depots in relation to the anticipated demand over the next few days.
And it was all trunked around the countryside overnight in great big trunkers from factory to
factory and then smaller. delivery vehicles from factory store to local depot.
And this damn programme, we never quite found out how to stop it printing out that it wanted you to
send one pallet of ice cream from a depot in Yorkshire to a depot in Lancashire across the
Pennines. And the problem was that it held a library of vehicles.
And because in the local depot you used forklift trucks, it would schedule. When you said to me,
what on earth are you doing moving a pallet around? It's not practical. It said, oh, well, yes, but
you can use a forklift truck. Now, of course, that didn't go down too well with the credibility of
the system amongst the operational management. You were telling them to send a forklift truck from
Yorkshire to Lancashire across the M62 to optimise the balance of polar maids between one depot and
another.
So we did make mistakes. And then that pattern was repeated throughout the Lyon's Empire.
So every one of these boxes is that sheet.
And this was the master plan that we were working on through the 60s and 70s. And it never came to
pass because events overtook us. It did. Events like...
Local VDUs for input and output made the turnaround document concept obsolete.
More subtly in terms of management philosophies, the individual operating companies developed a lot
of independence, they did.
And therefore, they started to set up their own computer departments. And as computers became
smaller and more powerful, each operating company invested in its own machine. Because, I mean, you
take something like T, which became Lyons Tetley, then that had a...
I mean, that was a £40 million turnover company in its own right.
And throughout... So you were talking about quite substantial businesses. and they understandably
developed divisional autonomy. So this master plan never actually got anywhere near completion.
But I show it to you to show how the philosophy that had led a board of directors to put £3,000
into a machine at Cambridge University continued.
But of course, J. Lyons went on an international expansion. Wimpy and fast food outlets instead of
tea shops, Baskin Robbins, Tetley Tea, Continental Foods. And because it was a family-owned firm,
it funded this expansion by loans rather than the issue of share capital.
And it then got hit by falling sterling values,
rising interest rates. and it was generating £20 million a year profit and paying 16 million quid
out servicing its debt. So a guy called Showering,
who'd made a lot of money out of baby sham and built up an empire called Allied Breweries,
took Lyons over and it became the food division of an outfit called Allied Lyons that subsequently
became Allied Demeck. And some years ago now, the Allied Demeck board decided to get out of food.
So all that was Lyons was sold off. And now you have a few brand names.
I'd like to think that J. Lyons would be remembered for LEO as well as tea shops. I would.
As I already said, there's not much material around.
But it would be sad if lessons of such an imaginative and innovative project were forgotten. So I
hope you've enjoyed what I've shared with you this evening. if you've got any questions.
[HOST] Well, thank you, Alan, for that most interesting insight into a subject that I suspect very few of
us knew anything about at all. So I hope it stimulated a few thoughts, and are there any questions,
please?
Stand, everybody. Yes, madam. You mentioned that the clerical staff were not made,
in today's terms, redundant. Where were they used? Were they spread around on these?
[AK] They were used to help the management implement what the output of the computer systems was telling
them. So effectively instead of pen pushing things,
they became management accountants instead of input clerks. So instead of calculating pay,
they were helping the management to work out where the business was running profitably and where it
wasn't. And there was natural wastage as well. So, I mean, as people retired or resigned,
they just weren't replaced.
But Lyons was a family firm, very paternalistic. And I remember the shockwaves that went through
the organisation. It would have been in about the 1970s when people were made redundant for the
first time because of business rationalisation, nothing to do with the computer system.
But I remember my father being quite pleased when I went home and told him I'd got this £700 a year
job with J. Lyons on computers, because he said, oh, you'll be all right there, son. He said, you
know, they're good people. And he was a rather blunt gentleman, so forgive me for repeating his
language, but he said, as long as you don't... pinch the petty cash twice or rape the director's
daughter and not marry her you'll be all right there for life but that was that was the ethos of
that sort of area large organizations were very paternalistic so they they looked after people and
they they genuinely said nobody would have made redundant because of this machine and it really was
used to run the business more efficiently which is why i gave you an indication of the early
applications wasn't about wasn't about getting rid of people
[SPEAKER 2] [Undetected question] [AK] Well the people that Thompson and Standingford were talking to in the States in 47 and there were
always arguments between them. Standingford left quite soon afterwards.
But Thompson used to say it was his flash of inspiration in a British Overseas Airways plane
halfway back across the Atlantic or something. To say, look, these devices will enable us to jump,
punch cards, which they quite rightly thought were abominations unto the Lord.
And they'd expressed some interest at, I think it was Harvard,
about the machine there that was used purely for scientific work and had inquired if there was
anything comparable. Did the Harvard chaps know if there was anything comparable? And they'd
mentioned things like Manchester and Cambridge. I don't know personally why Thompson went to
Cambridge. [SPEAKER 3] So it seems almost a parallel, it's a parallel, quite simple. [AK] Yes, oh yes,
yes, yes. [SPEAKER 3] Not much more modest. [AK] No, no, no, no, yeah.
But, you know, I mean, Maurice Wilkes at Cambridge was building the thing to calculate prime
numbers and factor this and he was.
And he wasn't dreadfully interested.
They did put an early version of payroll on EDSAC and it ran for 17 minutes and then stopped and
nobody ever found out why.
[SPEAKER 4] Observation, if I may, the thing that strikes me about your story of LEO, when it was invented,
when it was built, you had a very, very clear idea what you wanted to use it for.
Yes. Now, I remember the introduction of the mainframe computer, with BDUs and things like that,
with East Electricity. I remember Dent West, our training college,
and we had this, we got a distress in the audience. [Undetected audio]
[AK] It would be a polite word, wouldn't it?
Yes, [undetected audio]
yeah.
I said, why can't I ask the computer the question?
A list of all my jobs in detail, if I want that list.
If somebody's been booking my job number, I want to know who's done it. If nobody's been booking on
my job number, I don't want the 36 feet. [AK] But that's two threads I tried to put across to you.
In the 30s, John Simmons and his systems research department built that sort of thinking.
into all the lines, clerical admin operations, that you analyse and feed back to the management
information, not just piles of paper.
And the early LEO applications followed that same concept on bakery manufacture and the tea shops
ordering and the tea blending and all the other things. And one reason why LEO Computers was merged
off by Lyons was that when they were trying to sell LEOs to other people,
which they did with some success, they wanted as part of the sale to put their systems analysts
into the client business and help the client business. redesign its management systems first and
then computerise them. Now that meant LEO Computers Limited lost lots of orders because a good many
firms said, blow all of that, we just want a computer in a big glass window,
you know, in the front of our head office.
And they were quite happy for IBM, I'm sorry to be rude about IBM, we've got any IBM employees in
the room for you, sorry. Now you said you were Honeywell. Yes. Oh,
well, never mind. I mean, IBM will add Honeywell and Burroughs and the others were quite an ICT.
They were quite happy to supply computers, they were, and software and application packages.
And if it didn't actually do what the customer wanted, the customer had to alter his systems,
was the philosophy. And the people bought them because they'd all got IBM and Honeywell and Burrows
punch card installations and it was said that it would be easy to transfer them across one to the
other. So you ended up computerising a lot of not very good punch card based information systems.
I'm sorry you suffered, but if you'd bought a LEO 3 you wouldn't have done.
The magic of sales talk. No, no, no. I've sold ice cream and tea bags in my...
We've now got a new cabinet kit.
What we've got to do is to remember what we're in this company for and we have to learn how to use
it. We don't become computer programmers,
in my case, like electricity, and again, take the right off the ball as to what we're there for.
Well, thank you, John. John. Alan.
Again. Yes, indeed. It's the age, you know. Thanks once again.
Most interesting. I'd like to express your appreciation in the usual way. And just before you do,
the usual commercial. We shall be very pleased to see you at our next lecture on the 3rd of
February, which is John Walker. Another interesting subject, Britain's Most Secret Army.
Hope we shall see you then. Thank you.

Provenance :
Created by Alan King and donated to the LEO Society via Peter Byford

Copyright
The video above may be subject to copyright of the original author and is present only on this website for non-profit educational purposes only. If you are the original copyright owner and wish to have the media removed from this website please contact us using the relevant contact email address listed on our contact page.