Speculations on the Linking of Mechanical Computers

Some time in 1812 (or so the story goes) Charles Babbage (1791-1871) and John Herschel (later Sir John, the astronomer and physicist (1792-1871)) were sitting in Babbage's room in Cambridge checking some calculations which had been done for them, which they suspected contained many errors. "I wish to God these calculations had been executed by steam," remarked Babbage. "It is quite possible," replied Herschel, and so started Babbage thinking about the design of an automatic calculating machine.[1]

The British Government, which was greatly concerned by inaccuracies in existing tables, and the labour involved in preparing new ones, agreed to support a project to build a full-scale difference engine, which would compute six orders of difference, each of 20 places. The engine would not only compute the values of the function required, but would set them in type, thereby avoiding compositors' errors, which were even more numerous than computation and copying errors. The work started in 1823. It was an extraordinarily ambitious venture, undertaken fifty years before either typesetting machines or typewriters had been invented.[3]

For fourteen years (with a standstill of five years) Babbage worked on the design of his great machine at a cost to the country of £17,000. Gradually the Government lost patience, in spite of its eagerness for the machine, and Babbage himself became more and more engrossed in a still more ambitious dream: an "analytical engine" which would be a truly universal automatic calculating machine, instead of being limited to differencing.[4]

But what if Babbage had had the resources and personal commitment to make the computer (or "analytic engine") a reality in 1825? Such is the background history for the action in the science fiction novel The Difference Engine by William Gibson and Bruce Sterling. The novel takes place during the full flowering of the computer age but it is 1855. The analytic engine is a steam powered clockwork device of immense size.[7] Huge expenditures of money and manpower by the British Empire and other nations were required to bring Babbage's engines into existence. Using the engines, however, the governments can exercise more control over their populations than ever before in human history.

Although the novel never discusses the matter, to make the best use of the engines, perhaps it would be a good idea to link them together into some kind of network. Therefore, I will examine this question:

Given the limitations of 1850's technology, but in no way limited in money nor manpower, and using fully modern concepts of computers and networks, what would it take to make networks out of the computer "engines" as described in the novel The Difference Engine by Gibson and Sterling?

The novel The Difference Engine tells, in passing, of computers run by the government and others owned by businesses. One of the novel's main characters, Mallory, seeking information, goes to the Central Statistics Bureau in London. With the help of a guide, Tobias, Mallory finds the "engines" of the Quantitative Criminology department:

Behind the glass loomed a vast hall of towering Engines-so many that at first Mallory thought the walls must surely be lined with mirrors, like a fancy ballroom. It was like some carnival deception, meant to trick the eye-the giant identical Engines, clock-like constructions of intricately interlocking brass, big as rail-cars set on end, each on its foot-thick padded blocks. The white-washed ceiling, thirty feet overhead, was alive with spinning pulley-belts, the lesser gears drawing power from tremendous spoked flywheels on socketed iron columns. White-coated clackers [computer workers], dwarfed by their machines, paced the spotless aisles. Their hair was swaddled in wrinkled white berets, their mouths and noses hidden behind squared of white gauze.

Tobias glanced at these majestic racks of gearage with absolute indifference. "All day starin' at little holes. No mistakes, either! Hit a key-punch wrong and it's all the difference between a clergyman and an arsonist. Many's the poor innocent bastard ruined like that…."

The tick and sizzle of the monster clockwork muffled his words.[8]

He chose to place his bet with the thoroughly modern firm of Dwyer and Company, rather than the venerable and perhaps marginally more reputable Tattersall's. He had frequently passed Dwyer's brightly lit establishment in St. Martin's Lane, hearing the deep brassy whirring of the three Engines they employed.[9]

The side-by-side government Engines described above could easily be linked together to form a local area network by purely mechanical means. A rod could run through the row of Engines. It would turn, and/or be turned by the clock-work version of the network adapter within each Engine. This would be a bus topology. All the Engines on the rod could send or receive data over the rod. A mechanical CSMA system could be implemented whereby if the rod is already turning, the gears necessary to make it turn could not engage.

There are several ways a rod, by turning or spinning, can pass data. The best method would be dictated by the actual workings of the Engines themselves. Other than general hints, I was not able to locate detailed descriptions of the workings of Babbage's Engine in the small research I did for this project.

The degree of the turn of the rod could carry data: A turn of 90ºfrom the starting (or null) point could register as a 0 bit, while a turn of 90ºin the opposite direction from the starting (or null) point could register as a 1 bit.

Other methods might use a continuously spinning rod. The rod could then be treated as a carrier wave. Changes in the speed of rod rotation would be similar to amplitude modulation. The changes in speed would carry data. The changes in speed would be detected by a device not unlike the automatic transmission of an automobile.

The continuously spinning rod could abruptly be shifted into a new point on its rotation. This would be similar to phase shift modulation. Clockwork network interface equipment within the Engines would be synchronized with the rod to detect the shifts. Because the hardware would be able to measure the amount of shift in the rotating rod, phase shifts can encode more than one bit of data.[10]

David E. Hughes patented a system in 1855 to use a continuously spinning cylinder to send data over the telegraph lines. His system was extensively used between 1867 and 1932 in what is commonly called stock tickers (or ticker-tape printers).[11]

There are, however, some major drawbacks to using rods to make a network. If enough rods connect the Engines together, do they not become one big Engine instead of a network? Making one big Engine out of any number of side-by-side Engines may well be easier than dealing with network interface! Also, rods have severe flexibility and expandability problems. For rods to work well the Engines should be in a row and close together.

There was a technology available in the 1850's that could very well have been adapted to networking needs.

In 1831 Joseph Henry first successfully applied electromagnetic principles to telegraphy.[12] Wide public use of telegraphy began in 1844.[13]

For a business such as Dwyer and Company (cited in a quote earlier), telegraph technology offers a flexible and inexpensive alternative to the rod method of networking their three Engines. They may not have the space to put the huge Engines side-by-side. Telegraph wires could go where moving rods could not.

Because Dwyer and Company has only three Engines, a simple point-to-point system may be used. Duplex telegraphy, developed by Robert Walker in 1854[14], allows data to be sent in both directions simultaneously over a single telegraph wire.

In the 1850's, a 5-bit precursor to ASCII was developed. It was refined and put to use by Jean-Maurice-Émile Baudot. It is called the Baudot Code. It was used for over a hundred years to send international telegrams.[15] However, the 5-bit Baudot Code was inadequate for computer use. It was quickly replaced by 7-bit ASCII with the dawn of the computer age.[16] It can therefore be speculated that an ASCII-like code would be developed with the advent of mechanical computing Engines.

As with the electronic computer, the mechanical computing Engine could probably send and receive data faster than its clock-work CPU could process the data. A buffering system would be required. Buffering could be achieved by means of telegrapher's tape. Automatic telegraphy using rolls of paper was first developed by E. Davy in 1838.[17] Clock-work network interface equipment within the Engine would punch out the tape as fast as the data came in. Then it would read the data for processing at whatever speed was suitable for the mechanical CPU. (A direct reference to telegrapher's tape is made on page 349 of the Gibson & Sterling novel.)

For the three Engines of Dwyer and Company a point-to-point system may be all that is needed. To build a network with more Engines, however, could require the use of the equivalent of a bus topology. One of the concepts that would allow bus topology on a telegraph wire is time-division multiplexing. The idea of synchronized rotary motion at both circuit ends was proposed by Claude Chappe in 1790. Chappe's idea was made into a time-division multiplexing device in 1853 (It was later refined by Baudot).[18]

All Engines on the bus would have a synchronized clock-like rotating device in their network interface equipment. If there are five Engines on the bus, then one fifth of each rotation would be allotted to each Engine for the sending of data. Receiving data would be done by all Engines all the time. (Addressing procedures having been worked out.)

For larger networks, I see no reason why a token ring topology would not work well. In fact, I think a token ring system would work even with Engines connected by rod.

This paper in no way comes close to exhausting all possible ways to make networks out of mechanical analytical Engines. It is scarcely even a start. The linking would be a problem, but not impossible.

The key lesson of this project is that many of the concepts we think of as brand new and ultra modern have roots and antecedents that go back farther than one would think.