Journalists who want to write about practically anything now seem to begin, as a matter of course, in some remote, unlikely place, and in a sort of random walk approach the subject matter in question only when and if it suits them. Here is an example—followed by a short quiz. It—the example—is taken from the opening paragraph of the prologue to the book that I have every intention of reviewing as soon as I am done with this prologue.
All the way to the horizon in the last light, the sea was just degrees of gray, rolling and frothy on the surface. From the cockpit of a small white sloop—she was thirty-five feet long—the waves looked like hills coming up from behind, and most of the crew preferred not to glance at them. There were no other boats in sight, but off to the south for awhile they could see the reassuring outlines of the coast. Then it got dark. Running under shortened sails in front of the northeaster, the boat rocked one way, gave a thump, and then it rolled the other. The pots and pans in the galley clanged. A six-pack of beer, which someone had forgotten to stow away, slid back and forth across the cabin floor, over and over again. Sometime late that night, one of the crew raised a voice against the wind and asked, “What are we trying to prove?”
Now the quiz. The subject of this book is:
A. The America’s Cup race
B. The sperm whale
C. The design of a computer called the Eclipse MV/8000.
If you answered “C” you are right.
Tracy Kidder, the author of The Soul of a New Machine, the opening paragraph of which you have just read, is, I would guess, a young writer. He is also, to judge from this book, a very good writer, but he has taken an awful chance. For the “new journalism” to work we must believe essentially every word. As outrageous as Tom Wolfe’s astronauts may appear, we have to believe that if we had been there and seen what Wolfe is describing, they would have been like that. Once we allow ourselves to think that any of his account might have been made up, the whole thing loses its point, and evaporates. Anyone can make things up, but you have to be Wolfe to organize reality in the astonishing but usually credible ways he does. However, here is a young writer who has chosen to write about a subject—the design of computers—so far removed from the common experience that for most people it is almost fiction to start with, and who has chosen to do it in such a way that I, at least, put the book down in the middle and said to myself, “This is fantastically interesting but how do I know that he simply did not make all of this up?”
It is true that in the book there is a lot of solid technical detail about computers. But, so what? Even Arthur Hailey visited airports before writing Airport. There is no serious documentation in Mr. Kidder’s book. His idea of a reference is to write, “Business Week blamed de Castro’s style of management mainly, but appeared not to understand that style any better than anyone else.” Business Week, one would suppose, has issues and page numbers. Is it too much to ask to be told which issue of Business Week is being discussed? There are no photographs of either the machines or the people and no diagrams of the former by either the latter or the former. Indeed, if I hadn’t liked the book as much as I did, I think I would have quit in the middle and refused to go on until someone convinced me I should believe any of it. As it is, I finished the book and then decided to try to find out whether I should believe any of it.
As I have mentioned, Tracy Kidder’s book is about the design of a computing machine—what is known as a superminicomputer—a physically small and extremely powerful and relatively inexpensive computer whose characteristics would have been unthinkable, say, twenty years ago. This computer is now being manufactured by a Massachusetts company called Data General, a company that was founded when some former employees of the celebrated Digital Equipment Company, also located near Boston, defected under somewhat murky circumstances and founded their own computer business. The Digital Equipment Corporation (DEC) had itself been founded by some MIT engineers who deserted the Lincoln Laboratories at MIT in the 1960s to start DEC. Thus MIT begat DEC who begat Data General. Both DEC and Data General are in the business of manufacturing minicomputers, a field that was left open when IBM decided not to enter it. Thus it was really IBM who begat DEC et al. Indeed, as some wag has quipped, the computer business consists of IBM and the Seven Dwarfs. (Very recently IBM entered the minicomputer business with a somewhat less than spectacular splash.)
The heroes of this book are a group of young “systems hackers”—young engineers and programmers—many in their early twenties, who set out to beat DEC in the design of a superminicomputer to end all superminis. The leader of this remarkable band was an equally remarkable young man—then in his late thirties—a folk singer turned computer engineer named Tom West. Indeed the prologue of Mr. Kidder’s book is a description of a recreational sailing trip that West took in which, it appears, he stayed up for four days straight, causing a psychologist who had been on board to ask West if this was his idea of a vacation, where did he work?
When I decided to look into the reality of Mr. Kidder’s book I decided that the first thing I would do was to phone Data General in Westborough and ask to speak to Mr. West. I fully expected to be told that there was no Mr. West but, instead, I was switched to a secretary who said that Mr. West would probably not be in that day but would get my message when he did come in. The next day I decided to try to “buy” a computer, namely the very Eclipse MV/8000 described in the book, so I once again called Data General and this time instead of asking for West I asked for the computer. I was told to call the company’s sales office in New York, and when I did I was asked to leave my name and business affiliation. I left my university phone number and am still waiting to hear from a Data General salesperson. *
From this experience I have concluded that it is not as simple to buy a computer as one might have thought. On the other hand, I did hear from West. I explained to him what my problem with Mr. Kidder’s book was and West assured me that as far as he could tell essentially everything in the book was correct and, except for a few minor misquotes, took place just as Mr. Kidder said it did. “After all,” West said, “he couldn’t have made that all up.” Next time perhaps Tracy Kidder will provide a bibliography and a few references and photographs.
All that having been said, what were Mr. West and his colleagues up to and why was it so difficult? In the jargon of the trade they were trying to bring Data General into “32 bit-hood.” The basic unit of information stored in a computer is the so-called “bit.” This is simply the information that a “switch”—in a general sense—is on or off. In modern computer parlance an 8-bit unit—for example 01101000—is known as a “byte.” (Four bits are known as a “nybble.”) It is easy to persuade oneself that the total number of distinct bytes—that is, of all ways of making different 8-bit “words”—is 28 = 256. If a machine had to process information in 8-bit units its code could only address 256 distinct locations in its internal memory. This is too small for most computer applications so that the standard minicomputer available in hobby shops processes information in 16-bit units, which means it can address 28 = 65,536 memory locations. However, the large machines—for example, the IBM machines—use 32-bit words, which means that they can address 4,294,967,296 memory locations, and this memory space is essential for complex programs if they are to run efficiently.
In the late 1970s DEC produced a minicomputer called the VAX 11/780 which operated with 32-bit words. This meant that everyone in the field had to produce a competitive product. As West put it, according to Kidder. “We’re gonna get schmeared if we don’t react to VAX.” Data General wanted both a faster computer than the DEC VAX and also one that would be compatible with its own “software”—its previously written programs. In the fall of 1978 West put together his team of young designers whose objective was to design and “get out the door,” a 32-bit Data General product in less than a year—a record time for producing a new computer. In the event, the machine got out the door and into the Roosevelt Hotel in New York on April 29, 1980, where it made its first public appearance. By this time the unusual group of people that had built it had largely dispersed and West himself had moved up into the administration of Data General where, presumably, my phone call found him. Of the remarkable experience of putting together the machine that they came to love, West is quoted by Kidder as saying, “It was a summer romance. But that’s all right. Summer romances are some of the best things that ever happen.”
To design the machine West decided to use as a model the example of seymour Cray. Cray Computers are, at least in the scientific community, regarded as the Cadillacs of the computer industry, the state of the art. Cray’s company is located in the somewhat unlikely community of Chippewa Falls, Wisconsin. The company began trading on the New York Stock Exchange not long ago and when I last looked it had a price-earnings ratio of 47 and paid no dividends. Data General, on the same day, had a price-earnings ratio of 15 and also paid no dividends while IBM had a price-earnings ratio of 6.1 and paid dividends, which may go to show that Snow White may cry, in the words of Liberace, “all the way to the bank.”
Be that as it may, Cray had the inspired notion of hiring brilliant if inexperienced engineers right out of school on the grounds that they were too young to know what was impossible. He may also have realized that you had to pay them less, although, these days, a good systems hacker right out of college commands a starting salary which is something like the average salary of a full professor in physics at a major university. West liked the idea of hiring kids even though he realized that they might make the old hands look antiquated—that he might be hiring his own replacement. This, by the way, is one of the best arguments I know for the tenure system in universities. You can happily set about hiring people smarter than you are, secure in the knowledge that unless you are caught in an act of moral turpitude—a concept which in the present climate is so elusive as all but to defy definition—come what may, they can’t turn around and kick you out. There is no tenure in industry.
To find the right people the elders conducted personal examinations of the recruits resembling the ones that I read that Bernie Cornfeld used to give prospective mutual fund salesmen which turned around the question, “Do you sincerely want to be rich?”—meaning are you prepared to sell your grandmother. In this case, the question was, “Are you prepared to sell your soul to the machine for as long as it takes to get the job done?” If a recruit during the course of an interview said with the proper look in his eyes that he had used Data General machines and surely he could design a better one, he or she (they hired one female engineer) was a prime candidate for a job.
A case in point was the design of the simulator. With a simulator one can test the “intellectual capacity” of a machine design without actually building the hardware out of which the machine is composed. There is a general theorem credited to the great British mathematician and computer designer Alan Turing that says, in effect, that there is only one computer. Any general-purpose computer can be made in principle to do anything that any other one can do, although in practice one computer might take a week to do it and the other a fraction of a second.
Carl Alsing, who was thirty-five when West’s project began and was his second in command, came to the conclusion that they would need a program that could simulate the functions of their new machine. He knew that writing such a program usually took about a year and a half, by which time it would be too late to do their project any good. But he put together a team consisting of a veteran programmer, Dave Peck, who knew the job was probably hopeless, and a twenty-two-year-old wonder-kid named Neal Firth. He explained to Firth, in a general way, what the problem was and asked him how long it would take him to write the simulator. Firth allowed that it would probably take him at most two months. Kidder quotes Firth as saying, “I may be a little vain. I took a course in college in machine code. It was supposed to be a flunk-out course. To me, it was extremely simple. It’s always just seemed extremely logical to me, programming.” Six months later Firth and Peck had designed a perfectly operational simulator.
This story is typical both of the project and of the infectious spirit of the book. One gets caught up in the project. The machine almost becomes a living entity. One wants the darn thing to be built and one wants it to work. Not all of the young engineers could stand the pace. One of them, Josh Rosen, was too much of a perfectionist. He couldn’t stand the rough and ready way in which the project proceeded. He also didn’t much like the eighty-hour weeks with, by the way, no overtime. He once said, “If I spent only a sixty-hour week, I felt intensely guilty.” Eventually he simply burned out and one day he left the basement in Data General where the group worked, leaving a note on his computer terminal which read:
I’m going to a commune in Vermont and will deal with no unit of time shorter than a season.
Like most modern computers, the Eclipse MV/8000 has a cycle time—the basic clock in the machine—that functions in billionths of a second.
Since I was unable to “buy” an Eclipse MV/8000 I wasn’t sure what it was designed for—what the applications were supposed to be. But then, it would seem, neither were its designers. One of them commented to Kidder, “A sense of the applications is somewhat missing, but it doesn’t matter. We say the ultimate goal is to build a machine to run a multiprogramming reliability test. But I understand that people who buy computers do run other programs on them, like Adventure and Star Trek and things.”
In fact most people I know who have home computers use them to play games. Their children have learned to write programs that generate random numbers and the like. I watch them with awe. I am of a generation that grew up just before the computer revolution. Programming to me has always seemed like work—the sort of work I do for a living when I am not writing book reviews. I cannot imagine coming home from a day of teaching and doing physics to program a home computer to play “Maze Mouse”—“The mouse with a mind!”—or “Amoeba Killer”—“You command a one-man submarine that has been shrunken to the size of a bacteria.” But I find these machines absolutely fascinating, at least in part because they seem to provide models—albeit crude—of behavior we call intelligent.
In this respect the purely arithmetical capacity of these machines is the least of it. There does not seem to be any obvious limit to what they can be made to do and already they cause many people, myself included, to wonder what the definition of a “machine” really is. They seem to be defining our epoch. For readers who would like to know what it takes to make one, how they are organized, and who the people are who put them together, I strongly recommend Tracy Kidder’s book. I do not know anything quite like it. It tells a story far removed from our daily experience, and while it may seem implausible, it has the ring of truth.
October 8, 1981
A few days after I wrote the above I did get a phone call from the Data General branch sales office. I learned that the Eclipse MV/8000 was designed for “real-time applications” such as airline reservations. It costs between two hundred thousand and a million dollars, depending on what one wants with it in the way of auxiliary equipment. The company says it is doing nicely with them. ↩