Code_ The Hidden Language of Computer Hardware and Software - Charles Petzold [87]
We should forgive Hollerith for not implementing a binary system for recording census information: Converting an age to binary numbers was a little too much to ask of the 1890 census takers. There's also a practical reason why a system of punched cards can't be entirely binary. A binary system would produce cases in which all the holes (or nearly all) were punched, rendering the card very fragile and structurally unsound.
Census data is collected so that it can be counted, or tabulated. You want to know how many people live in each census district, of course, but it's also interesting to obtain information about the age distribution of the population. For this, Hollerith created a tabulating machine that combined hand operation and automation. An operator pressed a board containing 288 spring-loaded pins on each card. Pins corresponding to punched holes in the cards came into contact with a pool of mercury that completed an electrical circuit that triggered an electromagnet that incremented a decimal counter.
Hollerith also used electromagnets in a machine that sorted cards. For example, you might want to accumulate separate age statistics for each occupation that you've tallied. You first need to sort the cards by occupation and then accumulate the age statistics separately for each. The sorting machine used the same hand press as the tabulator, but the sorter had electromagnets to open a hatch to one of 26 separate compartments. The operator dropped the card into the compartment and manually closed the hatch.
This experiment in automating the 1890 census was a resounding success. All told, over 62 million cards were processed. They contained twice as much data as was accumulated in the 1880 census, and the data was processed in about one-third the time. Hollerith and his inventions became known around the world. In 1895, he even traveled to Moscow and succeeded in selling his equipment for use in the very first Russian census, which occurred in 1897.
Herman Hollerith also set in motion a long trail of events. In 1896, he founded the Tabulating Machine Company to lease and sell the punch-card equipment. By 1911, with the help of a couple of mergers, it had become the Computing-Tabulating-Recording Company, or C-T-R. By 1915, the president of C-T-R was Thomas J. Watson (1874–1956), who in 1924 changed the name of the company to International Business Machines Corporation, or IBM.
By 1928, the original 1890 census cards had evolved into the famous "do not spindle, fold, or mutilate" IBM cards, with 80 columns and 12 rows. They remained in active use for over 50 years, and even in their later years were sometimes referred to as Hollerith cards. I'll describe the legacy of these cards more in Chapters Chapter 20, Chapter 21, and Chapter 24.
Before we move on to the twentieth century, let's not leave the nineteenth century with too warped a view about that era. For obvious reasons, in this book I've been focusing most closely on inventions that are digital in nature. These include the telegraph, Braille, Babbage's engines, and the Hollerith card. When working with digital concepts and devices, you might find it easy to think that the whole world must be digital. But the nineteenth century is characterized more by discoveries and inventions that were decidedly not digital. Indeed, very little of the natural world that we experience through our senses is digital. It's instead mostly a continuum that can't be so easily quantified.
Although Hollerith used relays in his card tabulators and sorters, people didn't really begin building computers using relays—electromechanical computers, as they were eventually called—until the mid 1930s. The relays used in these machines were generally not telegraph relays, but instead were relays developed for the telephone system to control the routing of calls.
Those early relay computers were not like the relay computer that we built in the last chapter. (As we'll see, I based the design of that computer on microprocessors from the 1970s.) In particular,