The Soul of a New Machine

by Tracy Kidder

  Above all, Rasala wanted around him engineers who took an interest in the entire computer, not just in the parts that they had designed. He said that was what was needed to get Eagle out the door on time. He wanted the Hardy Boys to bind into a real team, and he spoke with evident frustration of engineers who were reluctant to work on boards that someone else had designed, who felt comfortable only when working on their own. Josh Rosen, he believed, was one, only one of those who felt this way. "Rosen designed the ALU and he wants to work on the ALU and nothing else," Rasala complained. "But I have a good confidence level on that board. It's not the first priority now. I need him elsewhere." Much later, Rasala would say, "I realize now that Josh wasn't as comfortable with the ALU as I was." Back then, however, his feelings were different. It seemed that he had given up trying to win Rosen over. When he spoke to Rosen, Rasala's voice had an edge to it, like the one it sometimes had when he was describing himself. He teased Rosen from time to time. "Your ALU is not fast enough to keep up with this blazing machine, Josh," he told him one night in the lab. Ra- sala often spoke that way to Hardy Boys. The difference was that the others usually answered him back in kind. Rosen, on this occasion, just turned away.

  The days of the debugging wore on. In March West said, referring strictly to the debugging, "Most of the fear is gone now." He was speaking only for himself, however. The team had passed through the first sharp fear. But they had designed the machine much too fast for prudence. It had features that none of the group had dealt with before. At this stage none of them dared claim to understand in detail how all the parts worked and fit together. Sufficient cause for worry about the debugging remained.

  Alsing thought that even under the best circumstances, several kinds of fears inevitably attend a debugging. One was the fear of "the big mistake," the one that would be discovered late in the game and would require a major redesign — and with it, perhaps, a fatal delay. There was "the flakey fear": that they had designed Eagle and were debugging it in such a way that it would never be reliable or easy to build in large numbers. Rasala had that one fairly well in hand. But there was also "the bogeyman fear." "Just something dark and nameless," said Alsing, "that the machine just won't ever work." West said: "It's the infinite page fault you didn't anticipate. The bogeyman is the space your mind can't comprehend."

  Once in a while, Alsing said, he suffered a form of this last anxiety. "Maybe these guys writing the microcode are really bullshit artists, so full of shit they don't even know they're full of shit. Maybe this whole thing is bankrupt." Such thoughts came to him rarely, usually at night. They vanished when the sun rose. Probably they stemmed from the fact that Alsing wasn't writing this code himself or reviewing it closely. Rasala once remarked, "Yeah, th,e further you get from doing it yourself, the more demons you see."

  Gradually, Rasala came to own (having acquired the title to it from West) main anxiety over the machine. Anxiety became just something he lived with, like a bad back. For a while, I engaged in a ritualized conversation with him.

  "How's the machine, Ed?"

  "Ahhhh, la machine," he would say. "Let's see. How is it?" He would pull his chair up to the chart of whatever debugging schedule was currently hanging on his cubicle wall, and usually wound up explaining why they had fallen behind this schedule, too.

  One day he answered the question differently. "The machine. It's what everyone calls it. That's the whole thing — to build the machine.'" When Rasala heard that phrase, he thought of a movie called Duel, which he had seen on television a couple of years before. In it, the hero is chased by a trailer truck, for no apparent reason. Throughout the movie, as Rasala remembered it, neither the hero nor the viewer ever gets to see the truck's driver, if there is one, but only the front end of the semi looming like a huge evil face in the hero's rearview mirror, always threatening to drive the hero off the road, always returning just when it seems that the hero has escaped from the thing at last. "My favorite movie!" Rasala said.

  His five-year-old son, the older of his two children, soon found out his father's weak spot, and now when he got angry at Rasala, the child would say, "I hope your machine breaks, Daddy."

  At night, when the last Hardy Boys got ready to go home, they would leave Coke and Gollum running one of the many diagnostic programs, a long list of tests that the machine had to pass before it could be called truly functional. Usually they'd leave the prototypes running a test that they had already mastered. At home, in the middle of the night, Rasala would sometimes come awake with a start. He would not be conscious of having dreamed. He would awaken and find himself wondering if one of the machines had stopped working for some new, unknown reason. Or he would wake up thinking about the latest failure, the one whose cause they'd been looking for a whole week and still hadn't found. The bogeyman — la machine — was there in his bedroom.

  THE WONDERFUL MICROMACHINES

  To almost everything they touched, the Microteam attached their prefix. The office that four of them shared, sitting virtually knee to knee, had a sign on the door that said The Micropit; the room in which they held their weekly meeting was the micro- conference room. They gave out microawards and Carl Alsing had his microporch. One of them owned a van, which became the microbus. That winter several of them would go out riding in it on Friday afternoons, when West held his own weekly meetings with his managers. Then, in the first warm days of spring, they created the outdoor microlounge, to which they now repaired on those Friday afternoons.

  Near one corner of the back of Building 14A/B, down a steep, man-made, earthern embankment, lay a narrow slice of woods. It must have been a pasture once. An old stone wall ran through it. Several flat rocks had been removed and propped against the wall in such a way as to fashion the crude seats and backs of micro- chairs. The sky promised rain, and the leaves not yet being out, the lounge lay in full view of the camera mounted on the corner of the roof high above, but the several Microkids who gathered there didn't seem to notice the weather or the camera. Their mood was altered. Reclining, they talked about computers and other social issues.

  "Our nightmare is that you keep getting faults."

  "We have to do this project on the fly."

  "Yeah, we're all flying."

  "No, if you page-fault you have to look for it, but if you can't find it you have to page-fault to find it and you can't page-fault until you find it. That can happen with a stack fault, too. It's a basic kind of crock."

  "Sure. There's always a way to kill a computer."

  "We should build a trap that'll let us in, for fun, at the end."

  "That's a nice concept, leaving a trapdoor for later, but a real purist would want to get into a machine built by someone else."

  "Well, there aren't too many machines in this place that we can't control."

  "What does it mean to understand a computer?' I inquired. "Knowing where all the electrons go?"

  "Maybe the electrons aren't in the computer at all. They just turn on the lights."

  "Electrons are mathematical abstractions. So who can talk about electrons? We speak about electrons loosely."

  Everybody laughed and laughed, until, it seemed, everyone forgot what they were laughing about. One Microkid started talking about a model of computer that was ten years old. He said, "They're really ancient." Uttered there, beside the crumbling stone wall, the statement sounded odd.

  "I've been here since seven," said one of the team, apropos of nothing.

  "I've been here since January," said another.

  A couple were discussing solar energy, the military-industrial complex and education. Engineering school was awfully specialized; meanwhile, the liberal arts were in decline. "Liberal arts, they're not economically viable," said one.

  "Does that mean you should restructure education or restructure society?" said the other.

  "You should restructure society. That's for sure."

  It was a pleasant afternoon. Looking up toward the TV camera, I noticed a large bird's nest
in a nearby tree. Trying to catch their mood, I said: "Look at that bird's nest. Who do you suppose lives there?"

  Right away, in a tone of voice expressing something like appreciation, one of the Microkids looked up at the nest, nodded judiciously, and declaimed, "Carl Alsing."

  Part of the Microteam displayed a tendency to keep odd hours and play zany games. You'd expect that. Compared to that of the Hardy Boys, who were now bound by necessary schedules to the actual machine, the Microteam's job had an air of the ephemeral. In the main, they could write their code when they pleased, as long as they did it on time, which is to say, very quickly. The code, the go-between for language and machinery, is in itself odd stuff— a witches' brew, many novices feel the first time they see it work. "It takes people with devious minds," Microkid Dave Keating contended. "Almost all of us fall into that category. We're not quite system bombers, but we're shaded that way somewhat."

  "I hire Wests. Alsing hires Alsings," said West. But both West and Alsing had chosen recruits with fine academic records, unlike their own. And in other respects Alsing had shown eclectic taste. He hired a woman — and women were scarce in hardware design— as well as a former rock-and-roll musician. He picked some shy ones and some much given to practical jokes. At least three of his choices would turn out to be, when given the chance, fine engineers both of hardware and of microcode. And when he had hired Chuck Holland a couple of years earlier — this and the fact that he trained Holland was one of Alsing's crucial contributions to the project, as it turned out — he did so largely because of a piece of sculpture. "I'm working on this thing," Holland had said, when in his interview, Alsing had asked him what he did for fun.

  "Tell me about it," Alsing had said. So Holland had, and by the time he had finished, Alsing had felt sure that Holland could write microcode. "It was my clue that he could make something intricate work."

  From a distance Holland's sculpture looked like a rectangular cage, and closer up, a gallery of spiders' webs. It stood roughly five feet high. Thin steel rods attached to each other at odd angles made up the four outer walls of the cage. The top made a sort of funnel, somewhat like a coin box on a bus — four sloping planes that narrowed down around a little hole. Rather shyly, Holland dropped a handful of silvery steel balls into this funnel, and the sculpture went to work.

  Within it, in puzzling shapes, lay a complex of tracks made of two thin steel rods. A ball dropped through the hole at the top. An entire section of the cage seemed to move. Inside the webbed structure a section of the labyrinthine track swung down; the ball dropped onto it; the seesawing piece of track rose back to its starting place, blocking the next ball for a moment. Then it swung down again. The second ball dropped. In a moment, the thing was full of steel balls and all of it was softly clanging, like a city awakening at dawn. It made you blink to watch it work. The balls, descending, seemed to cross each other's paths. Now one seemed to be in the lead, now another. They made sudden sharp turns, zipped from one side of the cage to the other, disappeared from view behind a maze of fretwork, traveling ever so slowly, and reappeared as silver blurs.

  "Do it again."

  In fact, there was really just one track and the balls went down it one by one. The track was also discontinuous. A ball rolled down one stretch of it, balancing between two rods, and the rods of the track widened just enough, at just the right spot, so that the ball would drop through from one section of track down onto another section. "There are a lot of fine tolerances here. The track got a little rusty once and then it wouldn't work," said Holland. *

  "At first I just wanted to make it work. It was an engineering feat for me. Then I decided I wanted to hide the track, but then I decided the track's kind of neat so I let some of it show. I wanted it to have some meaning. I have my own ideas of form. Planes and so on." He ran his hand across one section of the outer walls, as if stroking an animal's fur. "These wires cut this plane, and I just think that's really neat." It had taken Holland months to build this thing, taking it apart, putting it back together, taking it apart again.

  His father was an engineer; Chuck Holland had wielded screwdrivers and pliers for as long as he could remember. He had met his first computer in high school — and it was an old IBM, of course. Good-looking, trim, neat in his appearance, he had a wild-looking smile, which he did not often feel like showing during the season of Eagle. He did not approve of the way the group was assaulting diagnostics, and he pleaded, without success, for a more diplomatic approach. He seemed quiet, pleasant, shy — all in all somewhat like Alsing, easily overlooked — and he felt that he had been, thus far in his career.

  Holland organized the microcode for Eagle. Each microinstruction of a microprogram is a string of 75 units of high and low voltages, represented by O's and l's when written down. Holland divided this standard 75-bit string into standard chunks of several bits each, each chunk being called a field. Each field contains some number of unique combinations of O's and l's and affects discrete parts of the hardware. Each unique combination is what Alsing called a microverb. Holland and Ken Holberger invented definitions for every possible microverb. The result was a sort of dictionary. Adding versatility to its contents was the fact that certain microverbs could, for instance, say one thing to the ALU and something altogether different to the Microsequencer.

  The possibilities for creating microcode full of internal contradictions were virtually unlimited. Holland had to guard against that. He drew up rules — of grammar, as it were — to make sure that the code and hardware would fit together and to prevent each microprogram from interfering in some subtle way with any other. He saw to it that every microcoder consulted the others when making a change in a microverb. The microverbs and rules for their application he put into a book, called UINST. U stands for "micro" and INST for "instruction set." The Microteam called it their bible. The Hardy Boys called it "a Sears, Roebuck catalogue" and "the Microteam's wishbook."

  UINST became a battlefield. Its contents changed every week. Holland and his troops would make their changes and the Hardy Boys would look at UINST and say, "There's no way we can do this function and this function in hardware." The two sides would argue and work out those problems. Then the Microteam would discover something else that was hard to do in microcode, and deciding that it should be done in hardware, they would insert in the next issue of UINST this wish for a change. The Hardy Boys were on guard. They'd scan UINST carefully, looking for new Micro- team mischief, and finding this new item, declare, "No way we can do that in hardware." And it was back to bargaining again.

  Control seemed to be nowhere and everywhere at once. It was an almost tangible commodity, passed from hand to hand down the hierarchy of the group, and everyone got some. West gave Alsing responsibility for getting the microcode done on time. Retaining ultimate authority, Alsing let Holland assume almost complete technical command; and after establishing general rules and while keeping an eye on them, Holland gave each Microkid virtual reign over a portion of the code.

  There were hundreds of basic instructions to be encoded in thousands of microinstructions. All of the Microteam worked on recoding the old 16-bit Eclipse instructions. There were the microdiagnostics, or "the Wringers," as they called them. They worked at all hours, and sometimes when the Hardy Boys needed a new section of code in order to move on with the debugging, they worked around the clock. Every member of the Microteam knew weeks of intense pressure, which would gradually subside, and then rise again.

  Jon Blau said: "You have narrowed your field of vision to a small little world, trying to make it a vision of your own mind, and that's the kick — getting control of something. Maybe that for me describes the kick — that I can be totally in control of XSH zero. That's the name of a signal, which is the electrical embodiment of an idea. It was wrong. This morning I found an easy way to fix it." It did not always seem healthy to Blau. "When you're concentrating on that little world, you leave everything else out." But he had to admit, he was enjoying himself. "T
hat's the big kick," he said. "That the guys with the purse strings are trusting a bunch of kids to come up with the answer to VAX. That's what bowls me over, that they haven't just put us in a corner somewhere, doing nothing."

  The whole process looked strange, confusing, improbable, even to some in the Eclipse Group. But although the group as a whole was missing deadlines, the process seemed to be working. They were developing the hardware and microcode more or less in concert, and the Microteam was holding up its end. Although most of them were only beginners, when they delivered their code to the lab, they usually did so on time; and they made mistakes, of course, but far fewer than Alsing had expected.

  One main reason why they did not fall into an enthusiastic chaos was Holland himself. He helped Alsing choose the team. He organized their work. He reviewed all of it carefully. He helped mediate the battles over UINST. He wrote some large chunks of code himself. No one had more skill than Holland at the art of intricate construction, and no one worked harder than he. Nobody had ordered him to do all this. Alsing had made the opportunity available, and Holland had signed up. On the whole, he was happy in this work. "I never had more people around me who felt like they were friends," he said. He, too, felt excitement in the air. But he had some experience. He knew that credit did not always fall to those who deserved it. Would Alsing and others take credit for things that he, Chuck Holland, had done? The fear made him gloomy sometimes. He seemed to be girding himself for disappointment.

  Holland did not make all arrangements. Another source of the Microteam's strength was a tool they called the simulator. They relied upon this thing. They loved it. They did not like to think where they would have been without it. And it was Alsing who arranged for the simulator, in an odd sort of way.