Then he discovered that they were no longer willing to follow him blindly.
The revelation was staggering. He spent most of the retreat trying to inveigle them into a fresh start on a hardware and software system radically different from Alto/Smalltalk. "No biological organism can live in its own waste products," he exclaimed one evening. In earlier days that would have started them off on a thematic tour of Ideaspace and an exploration of the multifarious purposes of death and renewal. This time they took it as a threat to their own investment in a growing body of work, and turned him down.
"When Alan said to burn our disk packs it was Dan Ingalls who would have had to do it," recalled Diana Merry. "And Dan couldn't do it. There were too many bits on those disks he would have to recreate again, which made it very, very hard to let go. We lost the will to break it all apart. Alan finally had to realize it wasn't going to happen."
Smalltalk was no longer his system. He had started it, but once he turned it over to the "completers" like Ingalls and Adele Goldberg it had morphed into their own property. Ingalls was particularly determined to transform Smalltalk into a full-service programming language, the last thing Kay desired. Were it anyone else, he might have been able to keep control of the effort. But he could not fight Dan Ingalls, one of the few people in the world whose skill in his chosen field awed even Alan Kay. He had to let it go and admire the system for what it was, not what he wished it to be.
"Pajaro led to Smalltalk-76, which was two hundred times faster than Smalltalk-72," Kay said later, unable to avoid expressing admiration for Ingalls's finely crafted code, no matter how far it departed from his own goal. "But," he added wistfully, "no kid ever wrote any code for Smalltalk-76."
The 1976 offsite permanently changed the human ecology of Kays group. It was not a disaster, exactly, as he acknowledged later. There were no shouting matches or overt recriminations. They returned to PARC still friends and colleagues. "But the absolute cohesiveness of the first four years never rejelled," he recalled. There might still be bicycle runs to Rosatis in town for beer and brainstorming, but the dirill of biking back to PARC and implementing some unprecedented new idea on the spot had evaporated. To Kay the team had lost its balance. The idea of a Dynabook for Children had "dimmed out," overwhelmed by everyone's professional imperatives and their desire to elaborate on what were now, to him, old ideas.
Kay remained preoccupied with a lesson he had assimilated from Marshall McLuhan: Once humans shape their tools, they turn around and "reshape us." That was fine if the tools were the right ones, but he was unconvinced that Smalltalk fell into that category any longer. Within a few weeks of the Pajaro Dunes offsite he enticed Adele Goldberg and Larry Tesler, two who were still willing to follow him off on a tangent, into joining his quest to regain the simplicity initiative.
The result was the Notetaker.
As Kay first sketched it out early in 1976, the Notetaker would be compact enough to perch on the user's lap. Although a direct descendant of the Alto in its basic concept, it would jettison the Alto's hard-wiring-and- microcode architecture in favor of one using microprocessors, the new family of silicon-based integrated circuits being developed by Intel and others. In his first sketches Kay incorporated a number of innovative technologies that had not yet appeared in the marketplace—but nothing, he insisted, that would not be available within a few years.
One thing that shortly became clear was that in building the Note- taker, Kay’s group would be on its own. The days of turning to CSL for hardware help were past. Lampson, whose word on technical issues was paramount in that lab, regarded Kay's new project with icy disdain.
"Sometimes Alan isn't really in touch with reality," he said later. The Notetaker offended Lampson's doctrine of research priorities, which stated that one needed to look ahead of the curve, but not too far ahead. As a Time Machine the Notetaker was pitched so far into the future that Lampson could not believe it would teach PARC anything useful. Given the limitations of the new chips, the machine was shaping up to be smaller, slower, and dumber than anything they had ever built.
"I told them that within the limitations of the technology of today you will not be able to build anything interesting," Lampson recalled. "You'll be able to build a gadget that will work and it will be possible to program it. But you won't be able to make it do anything interesting because the technology's just too limiting. And that turned out to be absolutely correct."
To Kay, Goldberg, and Tesler this was just Butler being exceedingly subtle. Who was he to say what was "interesting"? If they could build a truly portable machine that had, say, fifty per cent of the Alto's functionality, or thirty per cent, or ten, would that not be "interesting"? In any case, he had made the same arguments about Dick Shoup s Superpaint being too far ahead of the curve. Well, the Systems Science Lab had given Shoup a refuge from CSL's cold contempt. If necessary they would steam ahead with the Notetaker by themselves.
But there was more to Lampson's dismissal of the Notetaker than his doubts about the design. At about the time Kay first broached his idea for a compact portable machine, CSL had come under the spell of an idea that amounted to its polar antithesis. While Kay was scorning biggerism, the Computer Science Lab was embracing it, in the form of a dream computer they called the Dorado.
Like the Notetaker, the Dorado claimed the Alto as its direct forebear. But the resemblance ended drere. The Dorado was to be the most ambitious computer PARC ever built. Where the Notetaker was to be deliberately modest and compact, Dorado would be fast, powerful, big, and noisy. Where the Notetaker turned out small enough to fit inside a suitcase, the Dorado was the size of an industrial refrigerator, with five fans for heat dispersion drat roared like "a 747 taking off."
That the Dorado was the product of rampant biggerism is evident from the way Thacker, its principal designer, described his earliest ambitions: "The original idea was that it would continue in the simple tradition of the Alto. I described it as sort of a 40-nanosecond Nova (that is, a Nova with a much faster clock).*
His plan was simply to build a machine that would enable him to test a new generation of chips that promised to be faster and more reliable than those he had used in MAXC and the Alto. But by 1976, when the project finally got under way, those modest goals were overwhelmed by the vaulting ambitions of his colleagues. When Thacker laid out the Dorado's preliminary schematic on an Alto running SIL, his program for automated circuit design, there was scarcely anvthing available to run on the Alto except SIL. By the time the first Dorado circuit boards came off a manufacturing line to be fitted into a seven-foot cabinet, the flowering of PARC technology had produced Mesa, a programming system so big it could burst the seams of any Alto in the building.
"The Dorado certainly got more complex than I had planned on," Thacker said ruefully years later. "I do think it was the second-system syndrome at work. You're successful and you say, 'I'll build something that's a little bit better.' Dorado may have been better, but it was
*This 40-nanosecond clock cycle translates to a processor speed of about 25 megahertz (i.e. 25 million processor cycles per second). Compare this to today's desktop computers, which range in speed from about 133 to as much as 450 MHz.
certainly a lot more complicated. It took five years to get working and there were several false starts."
The first of these occurred while he was still assigned to the Systems Development Division in the old Building 34 across the street from Coyote Hill. By then the Dorado had been designated to be the heart of a digital copier SDD was planning to build. The flaw in this plan, it quickly emerged, was that the new chips Thacker had been so eager to fit in his new design were a major headache to use. Employing a technology known as ECL, for "emitter-coupled logic," they were indeed much faster and less buggy than the TTL—"transistor-transistor logic"—chips they had used in the Alto. But they were also terrible power hogs and threw off huge volumes of heat, which required patching in an additional power so
urce to drive a fan. By the time Thacker finished his first-cut design of the Dorado processor, he knew he could never make it cheap enough for SDD to ship as a commercial product.
The labs regrouped. Thacker started over on a processor for the Star that would use the buggy old (but familiar) TTL chips. This evolved into the ill-fated Dolphin. Meanwhile, the Dorado program returned to the Computer Science Lab, which was immune to the ferocious pressures of shipment deadlines and commercial price points afflicting SDD. Everyone at CSL knew from the start, Lampson recalled later, that the Dorado would be "entirely impractical as a product." But if the commercial marketplace was not prepared to spend the money to get one, they certainly were. The Dorado would be the biggest and best Time Machine ever.
The man assigned to oversee what was sure to be a record-breaking engineering project was Severe Ornstein. A solemn and intense individual whose professional resume included critical roles in the development of the LINC with Wes Clark and the construction of the first IMPs for the ARPANET at Bolt, Beranek & Newman, Ornstein's black beard and beeding eyebrows gave him the stern mien of a biblical prophet but masked an artists temperament beneath. He was the son of professional musicians—as a Harvard undergraduate he had briefly dallied with the idea of taking a music degree before settling instead on geology. In any case, his prickly temperament fit well into CSLs unforgiving environment, where his barked "Nonsense!" became as familiar a hunting call as Chuck Thackers "Bullshit!" Although he had been recruited to CSL by Elkind, his stubborn and rigorous mind rapidly won over Bob Taylor, who soon invited him into his inner circle, the Greybeards.
Ornstein's long experience with quixotic hardware projects made him a natural to head up the Dorado effort, even if his tough-minded assessment of the job made his colleagues uneasy about the scale of the undertaking. "I said it would take two years and ten people," he recalled. "That was twice what anyone else was talking about."
One day Lampson took him aside for an attempt at jawboning. "Look, Severo, I know you're right," he said. "But if you tell people how long it'll take they'll never start it. You have to lie to them." One could almost imagine Ornstein's eruptive reply: "Nonsense!" In the event, his estimate was right on the money.
Building the Dorado presented new logistical issues compared with MAXC, which was physically a bigger machine but was not expected to be mass-produced, or the Alto, which was mass-produced but small. Since there was no room for an assembly line on Coyote Hill, Ornstein rented another building about a mile away on Hanover Street, which became known as the "Garage."
The CSL engineers' fixation on building the Dorado helped fuel the Notetaker team's inclination to go in the opposite direction. Given CSL's determination to pervert the Dynabook concept by building a machine bigger than the Alto, "it'll be a long time before we have the Dynabook," Goldberg said one day. "Let's do something that's between the Alto and the Dynabook."
In time she came to think of the Notetaker as an electronic notebook for kids to use in school. The idea was doubly ingenious: It not only gave them a paradigm to shoot for, but also established the machine's physical dimensions. "That told us it had to be light enough to carry around so the kids could use it to take notes in class, then bring it home and back to school," Tesler observed.
"Adele had in mind the eMate," he added, referring to a small school- oriented laptop Apple Computer manufactured years later which bore a striking resemblance to Kay's original Dynabook sketches.* "She knew it had to be somewhat heavier than the eMate, though she was hoping it wouldn't turn out to be what it did, which was forty-five pounds, heavier than the kid."
Between 1976 and 1978 the Dorado and the Notetaker projects proceeded along parallel but antithetical courses. The Dorado was so huge in scale that its sheer physical power sometimes overwhelmed the Garage's efforts at quality control.
Recalled one technician, "It was easy to set a circuit board on fire because you had this unlimited amount of current. We saw several just literally burn up. The fans were so powerful you couldn't see where the smoke was coming out. You could smell it, and you knew that there was something seriously wrong, but you couldn't tell. So you had to shut the machine off and pull the boards to find out."
One day, working with a partner, he spotted a wisp of smoke coming off a board and leaned into the machine to pinpoint its source. Suddenly a dozen little capacitors went off like incendiary bombs. The technicians hit the floor. When the devices ceased ricocheting off the walls, they got to their feet and gingerly eyeballed the errant board. The capacitors, they realized, had been installed backwards. They had been ticking away like tiny time bombs until the powerful current finally blew them and the board to pieces.
Kay's group, meanwhile, focused not on how to pump an incendiary current through their machine, but how to make it run adequately on an electrical trickle and with the smallest and lightest components available.
Doug Fairbairn, who had joined the effort as chief hardware designer, was aware that Intel, which had long provided PARC with memory chips, had just introduced a processor-on-a-chip, the 8086. (This was the precursor of Intel's x86/Pentium line of microprocessors, which today power most
*The eMate was a hit in the education market but a failure in the general market. Apple discontinued the model during its financial retrenchment in 1998.
personal computers.) With Tesler’s help, he worked out a design in which three 8086’s would serve as the brains of the entire machine. They ordered the first chips Intel produced off the production line and promptly discovered a bug in the product, much to the manufacturers dismay.
"They said, We just gave you the 8086 last week! How could you report a bug already?'" Tesler recalled. But Intel had not reckoned with PARC's do-it-yourself mentality. Years earlier the lab had purchased a rare million-dollar machine known as a Stitchweld, which could turn out printed circuit boards overnight from a digital schematic prepared on Thacker's SIL program. "It turned out that no one else using the 8086 had Stitch-welds. Everyone else was going dirough complicated board designs, so they wouldn't know for months if there was a bug. But at Xerox we gave them that feedback in a few days."
Cramming eveiything inside a portable case remained their biggest challenge, for they did not intend to skimp on any of the technical features that made PARC machines distinctive. The Notetaker was to have a custom-built display with a seven-inch diagonal measurement and a touch-sensitive screen (to substitute for the mouse); stereo audio speakers and a built-in microphone; 128,000 bytes of main memory; a rechargeable battery; and an Ethernet port.
The latter, in fact, proved to be a particular headache. There was no question of going without it, of course—PARC could no more turn out a non-networked computer than it could go back to using electric typewriters. But fitting a standard Ethernet board—now boasting more than eighty chips—into the Notetaker's cramped interior was equally out of the question.
One day Tesler crossed the street to SDD's quarters in Building 34 and laid the dilemma before Ethernet's inventor, Bob Metcalfe.
"Why does it take so many chips?" he asked.
Metcalfe patiently explained the function of every chip on the standard Ethernet board and why each was indispensable. Tesler countered that plenty of the circuitry could be discarded without undermining Ethernet’s basic operability. While everybody at PARC had their heads stuck in their high-performance systems, he told Metcalfe, a new world of computer design was taking shape on the outside. PARC was going to have to adjust.
Tesler’s views on the matter approached the heretical. He was referring to the hobbyist market, which was indeed exploding. The first annual West Coast Computer Faire, held in April 1977, had attracted thousands of young fanatics from all over the Bay Area. These were serious amateurs who built computers named Altair and PET out of kits ordered by mail, and gathered every weekend to swap shortcuts and software at gatherings like the Homebrew Computer Club.
They were enchanted with
computing's gadgetry as an end in itself, just as a previous generation had been with their ham radio sets. For many years yet their mindset would remain alien to those who had learned their computing at PARC. But Tesler, one of the few PARC engineers familiar with this niche, already saw they knew plenty that PARC would need to learn. They had found new ways to move functions out of hardware into software and to cut corners to save money and space. It wouldn't do to dismiss them as kids playing with toys: Their computers worked.
"I don't think the chips are all necessary on the Ethernet board," he told Metcalfe. "These PC guys make their computers so cheap because they go through all these tricks. We ought to start doing the same."
"That's them," Metcalfe replied. "Our boards have to work perfectly."
Dealers of Lightning Page 38