5 - UNIVAC
by BERNARD A HODSON
After my stint in aerospace I joined a computer manufacturer. While the title of this section is called UNIVAC the company selling them at the time was named Remington Rand, a US based company. They also manufactured electric shavers and other things. I joined the company at its Canadian Head office in Toronto, where one of the first large computers was installed, called UNIVAC I.
The first UNIVAC was a monstrous system, about three metres long, two metres wide and almost three metres high. You could actually walk inside it because most of the circuitry was on the outside walls, presumably to help with the cooling. It took enormous power, so much so that when it was switched on the initial surge dimmed all the lights in the neighbourhood. It was also susceptible to any major power surges that might occur locally, such as the switching on of heavy machinery.. It used water coolers which vented to the outside. At the time I joined the Company only two or three had been sold in Canada.
In the early days there were a variety of devices being tried for the storage of data. The UNIVAC used a large tank full of mercury, which had to be at a constant temperature. When the unit was switched on after being off for some time, it took fromt 15-20 minutes for the mercury to reach operating temperature. The technique for storage with mercury was that electrical signals would be converted to sound signals and sent through the mercury as sound waves. At the other end they would be converted back to electrical signals. By knowing the speed of sound in mercury you knew when any particular data might be available. Other devices at the time used nickel delay lines, which operated on a similar principle. The time to get the data while it was in the mercury (where it was unavailable for use) was called the 'latency time'.
Initially I sat in on a short course about programming the UNIVAC, which also included the logic of the machine. Almost all early computer courses covered the logical elements of the systems you were using. At this time we were still programming in a machine like language but a Commander in the US Navy, attached to the company head office in Philadelphia, named Grace Hopper, was attempting to develop an English like language for the programming of business applications. I attended a course given by her that covered the early experiments with this language. Initially it was called B0 but was changed to B1 because of the association of the initial name with body smells. It was a very symbol oriented language at the time but in later years was improved considerably to become the almost universal language for programming business applications, called COBOL (COmmon Business Oriented Language).
The next development was that I was sent on a course to learn about a smaller and more business oriented computer called the UNIVAC File Computer. It had an innovation which was not repeated in the industry for many years. This was a hardware round robin switch that enabled you to file your data on one of several segments of a file storage device. This was the last computer course I went on as I found that in general I knew more about computers than the person giving the course, particularly if the manual was at all decent, as they tended to be in the early days.
On my return to Toronto the emphasis for sales was on this computer but the problem was how to analyse a customer's needs at a reasonable cost. This analysis could be time consuming. A colleague and I sat down together and established a technique which was to be the company pattern for many future years. We called it the 'broad brush study'. This is illustrated in detail in my book "Modern Data Processing for Management' published by Macmillan in Canada and Cahner's in the USA. Briefly it is a technique to interview a selection of key managers within a company, plus a selection of the working group, the selection being made from an existing organisation chart in consultation with the senior managers.
All interviews take between 1.5 and 2.5 hours, and the total time for the 'broad brush study', with generation of a full feasibility report is 3 times the interview time It was a very effective way of doing things, one which I continued to use for many years after leaving Remington Rand. I got to the point where in a 1.5 to 2.5 hour interview I could know in depth the total work of anyone I interviewed, whether it was a medical doctor, a nurse, a clerk, a government bureaucrat or head of an Exploration Department in an oil company. After a series of such interviews I would know more about the company than most of the managers employed there. The Head of Exploration at Imperial Oil, after I had done a study there, said that all senior managers should be subject to the interrogations I conducted. If you do it right people will tell you anything. One person I interviewed at the Queen's Printer in Ottawa, enthusiastically told me all a computer might do for him, not realising that what he told me would essentially eliminate his job.
In Ontario, Canada, at that time, there was a great deal of new road construction, as new towns developed and existing towns expanded. One of the problems that government was interested in having solved was what was called the 'cut and fill problem'. As a road is being constructed it will sometime go through depressions, which need to be 'filled' with material to keep the road reasonably level, or hilly sections which need to be 'cut' through and the cuttings disposed of. For the 'fill' portion pits are created to 'borrow' material for filling in the depressions, and dumps have to be established for any material 'cut' from hilly sections. The cost of the road construction is made up in a significant part by the transport of material from the borrow pits and to the dump areas. The problem then was to develop a computer program to optimise this road construction operation. In dealing with sales to governments of the time this problem had continually to be addressed.
At that time lotteries in Canada were illegal. Nevertheless the Province of Quebec under its Premier, Maurice Duplessis, decided that a computer system might be ideal to handle a lottery they were proposing for the Province, the profits going to help finance Municipalities. A colleague and I worked out how a lottery could be done with a computer and then had to write a proposal. Being Quebec, and being Duplessis, the proposal had to be written in French. My colleague was competent in French but as yet there were no French words for much of the equipment we were proposing, so we had to improvise and create reasonably sounding words for things like magnetic tape and magnetic core. The proposal did not go forward due to the country's opposition to lotteries. As far as we know this was the first ever computer proposal in French for Canada.
There were three types of storage media for data at that time, the most common being cards with holes punched in them to represent data. There were two main types of card, those by IBM with 80 columns of rectangular holes, each column containing the coded equivalent of a character or number, and a card with round holes by Remington Rand, containing 90 columns of coded data. Data was carried around in boxes and there were various mechanical devices to sort and process these cards. If by chance a box of cards was dropped (a not infrequent occurrence) then the cards had to be resorted to get the data back in the correct sequence for processing. Data on these cards would be read in to the computer by special card readers.
The third media was magnetic tape, in which data was coded along the length of a reel of tape. A character or number would consist of seven or eight magnetised spots across the width of the tape. The early tapes were made of metal and were quite heavy. A device built by Remington Rand was the 'key to tape' machine, known as a Unityper. For some reason this device never did catch on but was nevertheless a useful creation. I suppose that one reason was that you could not see the data on the tape (which was also metal) whereas with a card, if you knew the code, you could read the data. As well there were card readers which printed the decoded characters at the top of the card.
I was then transferred to Ottawa as a technical support person for two salesmen. By this time another computer had been introduced called a UNIVAC 120. This, however , was not a computer with an internal memory such as the other UNIVAC systems but was programmed externally by wires. There were two large boards containing hundreds of holes. Some of these were designated for instructions, either for 20 or for 40 instructions, and a number were designated for numeric data. Other data was read from punched cards. In order to do an application you had to think of the sequence of instructions that might be needed, along with the data elements, and physically wire it. It meant that for every application you needed a different set of wired boards. Because of the limited amount of instructions you would often have to split the application into two or three runs, with intermediate results being passed from one run to the next, on cards.
This struck me as ridiculous and it saw the birth of what many years later became GENETIX. I figured that by designating the instruction set on cards, using the 90 columns available on a card, I could create a set of wires to generalise the computer operation. This proved feasible and I was not only able to do business applications but a whole variety of mathematical computations as well. It meant that there need be only one wired board, rather than one for each application, and that from here on in applications could be developed externally. The proof of the concept came when it was necessary to develop a proposal for Canada Post, who were interested in computerising a savings bank operation. Instead of needing two or three passes it could be done in one, much to the consternation of IBM, who had a similar device but needed two passes for their proposed operation.
I then made a decision which resulted in quite a change in direction in later years. I decided to give a course in programming and logic, in my own time, in Ottawa and arranged with the local Carleton University College (later to become Carleton University) to offer premises and advertise the course. I do not think I received any pay for this one. Some twelve people attended. I gave the course on the UNIVAC I and everyone, including me, seemed to enjoy the experience. I also taught the logic of the computer. Logic became less and less taught through the years as part of programming courses, but in the early days both IBM and UNIVAC offered their clients excellent manuals on the logic of their computers.
Jumping ahead two or three years (to when I worked for Imperial Oil) I was making a presentation to several senior managers on some proposed study, when the most senior manager asked me if I would tell them how a computer worked, what was the logic used in these systems. Wow, I said, do you know what you have asked me to do, but they were still interested. Fortunately most of them were involved with the technicalities of oil reservoirs, drilling and pipelines. I prepared what I thought was a good overview and they thoroughly appreciated knowing 'how a computer worked, not just what it could do'. I think it had a positive effect on my career potential within the company.