by Tracy Kidder
A microcoder called a question to Holberger. Holberger called back, "Don't know."
"There's a rumor," declaimed Rasala, still standing in the center of the room, "that they're impeaching Holberger, to take away his Honorary Microcoder's Award."
Guyer was at the console of Gallifrey. He was trying to find the cause of a "subtle" failure in the I/O system, and to this end was trying to develop a method to increase the frequency of failures.
"That's a waste of time," said Rasala.
"No, it's not," said Guyer, and he kept on working.
Holberger was inserting his tweezers into the board of a running machine again.
"Holberger's risking his job again," called Rasala.
Suddenly, just a moment later, Holberger and several micro- coders — about a third of the crowd in the lab — went charging out of the room.
"Where you goin'?"-said Rasala.
"Home," said Holberger.
And a moment after that, Chuck Holland came in carrying a "disk pack"; it looked rather like a helmet. Inside it lay the code that would, presumably, enable them to play Adventure on an Eagle. Holland installed the disk pack in Gallifrey's disk drive, and Rasala sat down at the console. He typed a little, using only two fingers. His movements looked tentative. He stopped and made a blubbing sound with his tongue. Guyer pulled up his chair. Wallach, Holland and one of the diagnostic programmers approached and leaned in over Rasala's shoulder. Everyone was giving Rasala advice. It was a big moment. If it worked, if Gallifrey now opened up Colossal Cave, it would be the first time, after more than a year and a half of hard, nervous labor, that an Eagle had done something more than run tests. The computer would at last exist.
The console scratched out a message.
"Whoops," said the diagnostic programmer, looking over Rasala's shoulder. "You roached the disk."
"I told ya," said Wallach. "Ya shoulda put write inhibit."
"Now we gotta run fix-up," said Chuck Holland. This took some time, and then the scene was repeated, all hands gathered around the console, Rasala at the controls. Rasala typed. The console scratched out a reply.
"Good-bye," said Rasala. "Goodbye!"
FATAL ERROR, read the message on the console.
Again, they ran "fix-up," and again the console replied, to Rasala's request to initiate Adventure, FATAL ERROR.
Rasala gave up. There would be no Adventure tonight; Eagle just wasn't ready. Rasala fed Gallifrey a diagnostic program that it had already passed, wrote up a sign, saying Do Not Disturb, which he placed on top of the console, and was on his way out the door, when the console started scratching out a message. So he went back. The machine had rejected its microcode. Rasala went to a filing cabinet and leaned against it. Then he took a deep breath and went back to the machine. He and Guyer made some adjustments and finally Gallifrey started running again.
"Why's it working now, all of a sudden?" I asked.
"We don't know," said Rasala. "We don't know everything about the machine."
Reaching for some comforting words, I said, "Well, at least some of it works."
"Yeah," said Guyer. "And someday we'll know which part."
Eagle was failing its Multiprogramming Reliability Test mysteriously. It was blowing away, crashing, going to never-never land, and falling off the end of the world after every four hours or so of smooth running.
"Machines somewhere in the agony of the last few bugs are very vulnerable," said Alsing. "The shouting starts about it. It'll never work, and so on. Managers and support groups start saying this. Hangers-on say, 'Gee, I thought you'd get it done a lot sooner.' That's when people start talking about redesigning the whole thing."
Alsing added, "Watch out for Tom now."
* * *
West sat in his office. "I'm thinking of throwing the kids out of the lab and going in there with Rasala and fix it. It's true. I don't understand all the details of that sucker, but I will, and I'll get it to work."
He called for Rasala one evening. "I want to go into the lab."
"Gimme a few more days," said Rasala.
On September 25 Rasala said, "As of this morning. Eagle ran Multiprogramming twelve hours overnight without failing."
Holberger said, "I know how West felt, but he couldn't have done a thing."
Back in his office West said: "It wasn't an empty threat. The game, of course, though, is that when I say I'm gonna go in there, they haul ass, because they assume it's gonna be some kind of trivial, dumb thing."
It had been such a problem, having to do with noise. But it wasn't quite over.
On October 4 most of the team clustered around Gallifrey's console, with Holberger at the controls this time, and the machine opened up Colossal Cave.
Holberger had never played Adventure before, and he wasn't about to take it on seriously now. He fiddled around in the style of a complete novice, didn't even meet the pirate or garner a treasure, and then shut the game down.
There was another game in progress that interested him more. They were running the Whetstones again.
They watched and waited most of that day. When it was all over, there was a victory. Now that Eagle was an almost fully functional machine, the figures were much better. They came out to within a few digits of what the designers had hoped for more than a year before. West's commandment was satisfied: Eagle was about ten percent faster than VAX on single precision — at least according to the published figures for VAX — and it was about twice as fast as the fastest Eclipse.
But Gallifrey, the lead machine now, still wasn't all there. It was running all the toughest diagnostics, but failing occasionally on some of the lower-level ones. The Hardy Boys would leap for their analyzers, run the test again, and the failure wouldn't happen.
"A flake."
But where was it?
On October 6 the vice president, Carl Carman, came down to the lab as usual, and they told him about the flakey.
Carman is a man of medium height, in his forties, fair-haired, with skin rather pink from the sun — all in all cherubic-looking. He smiles like Alsing, mysteriously.
The ALU was sitting outside Gallifrey's frame, on the extender Gallifrey was running a low-level program. Carman said, "Hmmmmm." He walked over to the computer and, to the engineers' horror, he grasped the ALU board by its edges and shook it. At that instant, Gallifrey failed.
They knew where the problem lay now. Guyer and Holberger and Rasala spent most of the next day replacing all the sockets that held the chips in the center of the ALU, and when they finished, the flakey was gone for good.
"Carman did it," said Holberger. "He got it to fail by beating it up."
They still had to prepare reams of documents. They had to run more tests. Software had to complete the vast, complex 32-bit system software. Now and then they would meet new crises, failures that usually but not always turned out to be flakey ones. They'd continue to bite their nails over the supply of PALs. Manufacturing would have to figure out how to build Eagles and the Eclipse Group would have to help. Jim Guyer still had to make the I/O system work right, and that would take a while, and in the mean time Rasala would go on saying what he had been for these many months — as if by naming the worst he could prevent it from happening: "There's still a small but finite chance that there won't be an Eagle." Nevertheless, they had reached one approximate end. "It's a computer," Rasala said.
On Monday, October 8, a maintenance crew came into the lab with a large dolly. They loaded Gallifrey Eagle onto it carefully. They wheeled it down the hall to the Software Department. Several of the Eclipse Group walked along, as escorts, and a few of them went out that night and hoisted a few beers. But there was no ceremony this time. A few days later, in the gloom of the Cain Ridge Saloon, they shoved together some tables and held another PAL Award ceremony. Standing to give the presentation, Rasala paused, and turning to the people sitting next to him, he said behind his hand, "It's just an excuse to go out drinkin'."
Back in January Ra
sala had said they would open champagne when an Eagle went to Software. They had drunk champagne before, however, at less important milestones. Besides, now that the time had really come, Rasala no longer felt like drinking champagne. Partly, he was tired, more tired than he had ever felt in his life. Partly, he had begun to feel an emptiness of purpose looming; he had lived with "la machine" so long that he could not easily imagine life without it. In this mood, other problems took on perhaps an exaggerated importance. Something was going wrong, though — not with the machine, not with the group exactly — but they were in trouble.
A month or so before Gallifrey was shipped to Software, I had been chatting with Rasala in the hallway outside his office when an executive turned the corner, walking toward us. "Now we stop talking," muttered Rasala. So we did. He did not resume until the person had passed. And on the evening that I had come to see Rasala play Adventure on Gallifrey — the time the machine just wasn't up to the task — the basement had seemed steamy with intrigue. Rasala had said, cryptically, that false rumors were running around to the effect that West was on his way out.
I stopped in to see West. He had gotten his hair cut — short, like a soldier's or a respectable businessman's — and he looked plucked. He looked wan. He looked terribly thin. His face glowed slightly when I asked about the rumors. Smiling slightly, he remarked (paraphrasing Mark Twain), "The rumors of my death are greatly exaggerated;"
West had said once, "Data General is a paranoid place, and the Eclipse Group does nothing to, uh, relieve that." I'd heard a story, no doubt apocryphal, about a man who while working at Data General became so fearful of his colleagues that he took to crawling under his desk in order to read his mail. I'd been having some small arguments with the company myself, and that evening I'd twisted them out of all proportion in my mind. Coming into the main lobby at Westborough, noticing the usual glances of the others sitting there and the receptionist's pleasant smile, I had the feeling that they knew something about me that I didn't know. This passed, but I felt lightened when I left the building that night with Rasala.
Rasala looked grim. I thought perhaps it was the aborted launch into Adventure, but he seemed to feel positively cheerful about that. "If it never failed, there wouldn't be any point in making it work, would there?" he said.
We went to a Burger King a few miles from the plant. While we were waiting in line, Rasala said to me, "You felt it tonight."
"What?"
"Your whole mood changed. You were relieved to get out of there."
I agreed that was so. Then I asked him if he ever felt glad to get out of there.
Rasala looked me over, up and down, as if to say it was a foolish question. "Once a night," he replied.
CANARDS
West's risky decision to use the new chips called PALs had some troublesome consequences. Months passed before Data General could be -sure of getting enough of those parts to manufacture Eagles. So the machine's public debut was put off, again and again, until the spring of 1980. As time went on, however, it became clear that West had made the right choice; PALs really were a chip of the future. Moreover, the delay gave the programmers time to create a much more impressive number of software options than usually accompanies brand-new machines. Diagnostics, meanwhile, had time to perfect a full set of microdiagnostic programs, which would help make Eagle easy to repair. There was time, too, for the Eclipse Group to refine the machine. Computers are sensitive things. With some makes, you dare not switch the boards of one with those of another supposedly identical machine. But you could switch boards among Eagles without worry by the time they were done.
A few weeks before the announcement, Eagle did come down with another evil-seeming bug. It took the Hardy Boys a long time to isolate it. They got close but could not fully identify it. They were stymied, until Holberger came up with a solution. It involved the addition of a single wire to a circuit He said he couldn't ex plain why the repair would work. He just knew that it would, bet it wouldn't and as a consequence wound up buying Holberger coffee for the next couple of weeks.
Holberger made his mysterious repair over in the Software Department's
area. Watching, one of the programmers declared that logic design certainly was a peculiar art. But, truly, the Hardy Boys had arrived at the state in which they could feel what was wrong with their machine. And there wasn't much wrong with it, by the time it finally got out the company's door.
In the months that followed Eagle's debut, it would become evident that the computer was probably going to be a big win, just as West had promised. Rumor had it that by early 1981 the dollar value of orders for Eagles represented more than ten percent of the value of all new orders for Data General equipment So it did seem that Eagle had arrived just in time to rejuvenate the upper end of the company's product line. Data General was already late entering the supermini market, and might have missed the market altogether if it hadn't been for Eagle, because by the spring of 1981 North Carolina still had not produced a machine.
Who had been the prime mover behind this success, this act of recovery? The company's system of management? or the team itself? or West, or Edson de Castro perhaps?
Engineers in the Eclipse Group who had been around for some years still referred to de Castro as "the Captain" or as "the Man in the corner office," often in tones of amazement, admiration and fear. Eventually, some of the survivors of the team would refer to him, forthrightly, as "God."
"De Castro's down here only as a presence," West said, "yet as a presence he's an iron hand." Another time, West remarked, "De Castro does this crazy, apparently bungled stuff, but after a while you see this incredible order." And he told me: "It was de Castro who said, 'No mode bit.' It turns out that in a very succinct way he was describing the perfect machine." A succinct description indeed, if it was a description at all. West seemed to be suggesting that de Castro might have orchestrated from afar the the entire project, even including West's own conviction that Eagle's creators were working largely on their own. "I think we all have that feeling about de Castro," said a member of the team. He was only partly joking. Several engineers told me how sharp de Castro was on technical issues. Come to him with a plan that has a weak spot in it, a weak spot that you think you've cleverly disguised, they said, and he'll find it every time. You didn't put one over on de Castro. Things weren't done behind his back. Never mind how big the company had gotten, he was still in control.
Four straight-backed vinyl-upholstered chairs face de Castro's desk, which is absolutely clean, save for one small stack of papers with their edges squared. De Castro sits at a slightly lower elevation than his visitors, and when he sits down behind his desk, most of him disappears. He becomes, as it were, a sphinx — half desk, half man. He is thin and balding and wears on this occasion neither jacket nor tie. Asked about the meaning of the term competition for resources, he smiles broadly and says: "It's there. What it really does, in a sense, is allows for the accomplishment of certain projects that some people would prefer not to do."
Had the Eagle project always interested him or had it grown in importance gradually?
"From the start it was a very important project."
Was he pleased with the work of the Eclipse Group?
"Absolutely!" His voice falls. "They did a hell of a job."
But some members of the team felt that they had been rather neglected by the company.
"That doesn't surprise me," he says. 'That's frequently the case. There's often a conflict in people's minds. How much direction do they want?"
The team seemed to think of itself as an independent, entrepreneurial outfit within the company. Did that happen by accident or had he tried to foster that feeling?
De Castro has turned his head toward a wall during a lull in the conversation. Across his face there seems to flicker a look of wariness. He turns back, smiling again. "I think that if you try to oster that attitude, you will be unsuccessful I think that from our point of view we try to provide an environm
ent where those
dings oughta happen." His voice falls, and he adds, "In cases where they oughta happen."
Maybe de Castro orchestrated the project, or maybe, having put all the parts in place, he just let it happen — which might come to the same thing. Speaking privately, a veteran engineer bristled at that idea. "De Castro isn't that good," he said. "He's lucky and he's smart, but most importantly, he has people around him like West, who pull his ass out of a fire and then attribute it all to him." Members of the team had their own interpretations.
A month after Gallifrey Eagle was rolled slowly down the corridor to Software, Ed Rasala and several of the Hardy Boys got together at the Cain Ridge Saloon and did some reminiscing — disputatiously, as usual. At one point, Jim Guyer said: "We didn't get our commitment to this project from de Castro or Caiman or West. We got it from within ourselves. Nobody told us we had to put extra effort into the project."
Ken Holberger burst out laughing.
Guyer raised his voice. "We got it from within ourselves to put extra effort in the project."
Laughing hard, Holberger managed to blurt out, "Their idea was piped into our minds!"
"The company didn't ask for this machine," cried Guyer. "We gave it to them. We created that design."
Others raised their voices. Quietly, Rasala said, "West created that design."
Rasala's big forearms rested on the table, surrounding a mug of beer. I thought perhaps I had not heard him right. He had always insisted that Eagle belonged equally to every member of the team. There was a whiff of heresy in the air.
"What did you say that West created?"
"Eagle," said Rasala.
By then the others had stopped arguing and had turned toward Rasala, who was wearing one of his looks; it seemed to warn against contradiction.
