Jacquards' Web (24 page)

The counter worked in this way: when a needle ran through a hole in the card, it would dip into a little cup containing mer-cury. Mercury is an extremely good conductor of electricity.

Mercury also has the remarkable property of being the only metal that is liquid at a comfortable room temperature. This makes it useful as an electrical contact where the contact needs to 167

Jacquard’s Web

provide some flexibility as far as positioning is concerned; the flexibility that a liquid conductor can easily furnish.

In fact, there was no reason why the needle tip should not have made contact with any conductive metallic surface. But Hollerith was working at a time when electricity was not available from a main supply. Because of this, he needed to rely on lead-based accumulator batteries to operate his machines. These only provided a weak voltage. This made it essential to keep electrical resistance to a minimum.

Mercury provided the high-quality contact he needed, and does not readily corrode. Once the tip of the needle had dipped into the cup the circuit would be completed and the dial linked to that particular needle would advance by one increment. Pressing individual cards against the needles counted significant pieces of information gathered for the Census.

Like Babbage, Hollerith took care to build safeguards into his invention to avoid mistakes.

One problem that obviously had to be prevented was the danger of the punched card being entered into the system the wrong way up. To stop this occurring, every card was made with the same corner cut off to tally with a corresponding diag-onally shaped place in the card slot. If the card was the wrong way round, the corner would not fit neatly into the card slot and the card could not be laid in its place at all. Hollerith also took steps to avoid obvious card-punching errors. He arranged for the system’s wiring to deliver an audible alarm signal—usually a bell—to the operator if two mutually exclusive parameters (male and female, for example) were triggered simultaneously.

As well as this, it was useful for punched cards to be sorted according to the information they contained. The machine that did this was known, reasonably enough, as the
sorter
. Operators could sort cards into different boxes depending on the different criteria being investigated. If, say, the cards were being sorted according to the ages of the people each card represented, the 168

The first Jacquard looms that wove information

different boxes in the sorter might contain cards representing the age ranges
0
–
10
,
11
–
20
,
21
–
30
, and so on, with the final box containing cards for people older than
90
. The card-sorting process was controlled by each dial being connected to a switch that opened the lid of a small box whenever the dial advanced by one increment. The operator who had pressed the card against the needle switches would see the lid of one box open and would drop the card into that box.

These processes of counting and sorting cards might seem very slow and cumbersome from the description given. However, operators could become surprisingly adept at working with this equipment. Skilled operatives could count and sort several thousand cards every day. All the same, we can be fairly certain none of these operatives would have wished their working day to be a moment longer than it was.

Like many inventors, Hollerith was never entirely satisfied with his current range of machines. He was continually refining and improving them. The pantographic punch was eventually superseded by keyboard-driven punches. Once these were invented there was not much more scope for improving the devices that actually punched the cards. Hollerith could never do away with the requirement to punch in the raw data. Someone needed to do this laborious data-entering activity. In practice, Hollerith’s entire system relied on legions of card-punch operatives. He did, however, eventually develop an electric punch that needed only the lightest pressure on the key in order to activate it.

The fundamental principle that basic data needs to be entered manually still holds true in principle today. No matter how sophisticated a computer system might be, somebody, some-where, has to enter into the system the raw information—numbers, words, or other variables—that are being harvested in the first place. Modern optical scanners take some of the burden away, but in practice the manual entering of data is still extensively required for most applications. What is true of data is also 169

Jacquard’s Web

true of the often very large numbers of lines of computer programs. These lines are known within the computer industry as

‘code’. A huge computer program writing industry has sprung up in India, for example, where so-called ‘software factories’

employ large numbers of talented Indian computer programmers who (whether rightly or wrongly) can be paid a good deal less than their counterparts in the United States. These programmers write many of the major programs—often requiring millions of lines of code—that power standard office and domestic desktop computers and laptops.

With little choice but to enter the data (i.e. punch the cards) manually, Hollerith saw that the real scope for improvement of his machines lay in the counter and the sorter. In particular, he realized that it should be possible to develop a counter that would automatically ‘read’ each card without an operator having to press it manually against the grid. Here again one sees an intimate relationship to Jacquard’s conception of the punched cards being linked in a chain and momentarily locked and presented against the needle-grid automatically instead of being pushed in place by a manual operator.

Hollerith’s efforts to improve his technology were intensified as a result of the competition he encountered from a rival in the design and manufacture of punched-card machines. This man, James Powers, was also a brilliant engineer who made many important innovations. The actual invention of the
automatic
sorter—whereby the sorting process took place without any human intervention at all—is generally ascribed to Powers rather than to Hollerith, although competition between the two men was so quick-fire and fierce that establishing priority between them over their most important inventions is far from easy. But in general, with some exceptions, Powers always lagged a little way behind Hollerith.

By around
1900
, Hollerith had created a new system that made extensive use of automatic machines. The cards were first placed in a hopper that compressed them with a carefully 170

The first Jacquard looms that wove information

weighted block to ensure that only one punched card was released from the hopper at a time. The individual cards were now capable of being processed faster than the eye could see.

Each card was extracted automatically from the pack, pushed against tiny rods that sensed which holes were punched and which were not, and then dropped into one of various boxes depending on the results the tiny rods had found when pressed against the card. This process was basically analogous to the operation of the Jacquard loom; the difference was that the result of the cards being pressed against the mass of rods was that the cards themselves were sorted into different categories rather than that certain warp threads were raised to form a line of weaving.

Hollerith also introduced major improvements in the counting and sorting procedure. He had initially used clock-style dials to allow the operator to count the different numbers of cards in the different boxes, but soon found that it was easier for the operator to count accurately if the clock faces were replaced by cogwheels that allowed the user to read off the total number of features counted from a little ‘window’ in the machine. Eventually, as we saw above, the sorting machines did not even need manual operators to drop the cards into boxes. The cards moved over rollers, a sort of miniature conveyor belt, into whatever box had been triggered by the switches.

Hollerith felt that his new generation of automatic punched-card sorters and readers needed a new name. Around
1900
he started calling his automatic machines ‘tabulators’. He hired the most talented engineers of his age to keep improving the pioneering innovations in what was rapidly turning into a completely new industry.

An important additional development in
1907
saw the tabulators being fitted with ingenious electromagnetic printing mechanisms so that the user would be presented with a printed document incorporating the totals from each operation. Fitted with printing mechanisms, automatic tabulators were transformed into enormously powerful automatic tools for informa-171

Jacquard’s Web

tion handling and analysis. Indeed, the tabulator went on to greater and greater success, with punched-card systems forming the basis of important and popular information handling resources that remained in use throughout the first six decades of the twentieth century. Further improvements in the machines greatly increased their speed, accuracy, and information handling capacity. Even as late as
1960
, the data processing giant IBM was still earning more annual revenue from punched-card systems than it was from computers.

Today, one of the most curious aspects of the history of computing is how comprehensively the importance of tabulators has been forgotten. They were electromechanical digital machines that transformed the business of information handling in the first half of the twentieth century, yet today tabulators are only to be found in a few science museums around the world.

By the time Herman Hollerith went into business with his tabulators, the American economy was also being transformed by other ingenious office machines. Two of the most important were the typewriter and the cash register. The technology of these machines did not embody an enormously subtle and momentous idea as did the Jacquard loom, Babbage’s Analytical Engine, and Hollerith’s machines. All the same, the quality of their mechanical engineering was remarkable.

Hollerith’s inventions belonged to the same new wave of business machines. Yet there was an important difference in the nature of the market for tabulators compared to the one for other machines. Typewriters and cash registers were useful in almost every type of business and in every area of industry and commerce. Hollerith’s tabulation machines, on the other hand, at least at first, seemed to have only one application: census-taking.

It was a precarious basis for a business. After all, the US Census, by definition, only came round once a decade.

172

The first Jacquard looms that wove information

There was another important difference, too. Mass-market business machines were manufactured and sold by highly focused organizations that emphasized sales success and the meeting of customer needs. Hollerith, on the other hand, operated in a far more informal and even academic way. He had fanatical commitment to the technical quality of his machines and immense talents as an engineer and inventor. All the same, Hollerith essentially ran his company as a research and development organization.

For Hollerith, commercial success came late. His company only became a global force in the business machine world when control passed to a more commercially minded entrepreneur.

Between
1884
, when Hollerith filed his first patent for a census tabulation machine, and
1911
, when he sold the company, Hollerith built up a business that was remarkable more for the level of innovation it brought to its product development than for its commercial acumen and the dynamism of its sales force.

Another problem that handicapped Hollerith was that he tended to hire out his machines rather than selling them outright.

He had little alternative: his customers were governments that only required the machines for the duration of the Census. Governments had no motivation to buy the machines for permanent use. At first, when the machines were still being perfected technically, renting them out was not entirely commercially satisfactory. Their complexity meant that they had to be repaired and fitted with replacement parts regularly, and Hollerith was responsible for this and had to swallow the cost.

Later, however, Hollerith found that hiring out machines that were technically reliable could, after all, be a good commercial ruse. Each machine might, over a period of years, earn back its purchase price many times. The fact that he tended in most cases to take rental machines back once they were obsolete and then melted them down helps to explain why few tabulators survive today.

Hollerith also insisted that his government customers purchase his blank punched cards, a reasonable enough insistence 173

Jacquard’s Web

since punched cards could no more be re-used than could franked stamps. Some of Hollerith’s customers tried to economize here by manufacturing their own cards or buying them from rival manufacturers, but the machines needed a high-quality card to run smoothly, and customers generally accepted the need to buy Hollerith’s, which were of a better quality than any produced by his rivals.

Hollerith’s first major success was the
1890
United States Census, when his machines were employed to automate the entire counting process. The
1890
Census, which had looked so forbidding when viewed from the perspective of
1880
—when only manual means were available to handle the findings—was thoroughly tamed by the brilliance, efficiency, and practical advantages of Hollerith’s first generation of machines. The whole event was a triumph for the perfectionist inventor; the machines worked so effectively that they made the previous, manual, means of handling a census look primeval. The US Government reckoned to have saved more than two years’ work and $
5
million of tax-payer’s money, a prodigious sum by the standards of the time.

But just as important as the cost saving was the fact that the Hollerith machines made it possible to draw the widest range of information from the Census data. The most complicated tables showing relationships between different variables could be produced at no greater cost than the simplest. As for the accuracy and rapidity of the machines, everyone involved with the
1890