Kay's own 1969 thesis incorporated these ideas and others into what must be one of the oddest dissertations ever submitted for a doctorate in a scientific discipline, featuring as it did epigraphs from, among others, W. H. Auden, J. S. Bach, and Kahlil Gibran ("You would touch with your fingers the naked body of your dreams"). The hand-drawn illustrations included not only complex diagrams of functions and logical trees but line drawings of fanciful single-user machines. These had screen, keyboard, and mouse unified into a desktop console, a big brother to the portable all-purpose computer that had provoked such controversy when he described it at the ARPA grad students' conference two years earlier.
Kay's thesis outlined an interactive computer called the FLEX machine which he had designed in partnership with an unsung hardware genius named Ed Cheadle, who was an important engineer for a Salt Lake aerospace company. The FLEX incorporated many of the ideas Kay would develop in the coming years at PARC, including compactness, object-oriented programming, and the use of a display screen. But it was not quite the personal computer he envisioned, in part because it was not powerful enough to perform all the functions required by his ideal and in part because it utilized a complicated and stilted language which, as Kay recalled, "users found repellent to learn."
Despite its idiosyncrasies (or because of them), Kay’s thesis readily passed the muster of a five-man committee that included Sutherland and Evans. But he was tormented by a sense of things half-done. Parts of his FLEX machine could be implemented on existing hardware, but a truly suitable technology seemed to be tantalizingly just out of reach. "The big whammy for me came during a conference tour of the University of Illinois, where I saw a one-inch-square lump of glass and neon gas in which individual spots would light up on command—it was the first fiat-panel display. I spent the rest of the conference calculating just when the silicon of the FLEX could be put on the back of the display." The answer, according to Moore's Law, seemed to be at least ten years off.
If contemporary machines were inadequate, Kay’s goals had not changed. The quest was still for something simple enough for a child to use yet powerful enough to slake the human thirst for creativity. Kay imagined an invention called the "KiddiComp" or "Dynabook." To make the abstraction tangible, he built himself a model box about nine inches by twelve and a half inches deep, with a flat screen and keyboard drawn on the top surface, and filled it with lead pellets as a way of divining its optimum weight (about two pounds, he judged).
He was at loose ends, depressed over having failed to make his great idea materialize in more than cardboard form. While holding a temporary appointment at the Stanford artificial intelligence lab he underwent a year of gestalt therapy ("a very California thing to do"). He was on the verge of accepting a post at Carnegie-Mellon when Bob Taylor called him with the electrifying news that Lampson, Thacker, and several of their BCC colleagues were joining Xerox PARC en masse. Kay reconsidered his plans. Butler Lampson was one of his intellectual heroes. Through the ARPA grad student conferences he knew the others as first-class talents. If all these people were to converge at PARC under Bob Taylor, there was no telling what they could accomplish— even build his Dynabook.
One night he and Taylor stayed up nearly until dawn, batting around the possibilities implicit in a conjuncture of Xerox's money, Kay's ideas, and the engineering of Lampson and Thacker. A computer simple enough to be worked by children! Small enough to be carried under your arm! Powerful enough to drive a display in full color! There was only one thing, Taylor informed him at some point. Kay would not be assigned with the others to his lab, but to the competing Systems Science Lab under Bill Gunning.
Much has been made of Taylor's motives in keeping Kay out of his own lab. Some believe Taylor wished to place a "ringer" in the rival SSL—a "colonization," Kay said, "so two of the four labs at PARC would have ex-ARPA people" to more effectively propagate his ideas throughout the organization. It is just as likely that Taylors BCC coup had filled CSL's allotted head count for the moment. Since he was advising Gunning on recruitment anyway, there was nothing untoward in offering him Alan Kay.
It is also certain that landing in Gunning's SSL was Kay's lucky break. He tended to work as a loner—either by himself or as leader of a small team. Could he have maintained his intellectual autonomy in CSL, where the only group was Taylor's and the intellectual engine was Butler Lampson? Working out of SSL allowed Kay to work as a full participant in CSL's program without ceding his independent spirit. He could interact with CSL as a privileged equal, outside Taylor's direct supervision and Lampson's intellectual domination.
As events unfolded, Kay and Taylor apart proved more powerful a force than they would have been together. Where Taylor could be vague and inarticulate in describing computing's future, Kay was never less than crystal-clear. The day he came to PARC for his job interview, Rick Jones invited him into his office and asked him a stock question.
"What do you think your greatest achievement will be at PARC?" he asked.
"It'll be a personal computer," Kay replied.
"What's that?"
Spying a flat portfolio on Jones's desk the size of a student’s notebook, Kay seized it and flipped it open. "This will be a flat-panel display," he said, indicating the cover, which he held upright. "There'll be a keyboard here on the bottom, and enough power to store your mail, files, music, artwork, and books. All in a package about this size and weighing a couple of pounds. That's what I'm talking about."
He walked out, leaving Jones scratching his head and saying to himself, "Yeah, right."
With Kay's arrival the computer research team at PARC achieved critical mass. They had the people and the leadership, a seemingly unlimited amount of money, and Xerox's liberal commission to pursue whatever course of inquiry they wished.
All they needed now to start work was a computer. Pake gave them a month or so to study the available alternatives before recommending a system to be used by the entire research center. But in making their choice they provoked the first great donnybrook of PARC's young existence.
CHAPTER 7
The Clone
Perhaps Taylor and his hand-picked team should have known by sheer intuition that when you are about to spend a half-million dollars of your employers money, you do not go spending it on a competitors product. After all, Ford employees do not drive Toyotas, and the soda machines at Pepsi headquarters do not get stocked with Coke.
Or perhaps, having almost all come from academia and government service, they were simply too green at working for an industrial enterprise to understand why Xerox's computer company, SDS, would take amiss Taylor's proposal that PARC purchase as its main computer a machine manufactured by its commercial archenemy. Especially when the choice seemed to them a no-brainer.
Yet that is what happened. As Taylor put it later, "We upset the apple cart and didn't realize how badly." Badly enough: PARC's opening blunder would not be forgotten at Xerox headquarters for many years. *
The Computer System Lab's inaugural task of selecting the main computing hardware for all the PARC labs seemed straightforward enough from the standpoint of pure engineering. When the lab drew up the specifications there was no need to hold a colloquium to weigh their options. There was only one: The PDP-10, manufactured by Digital Equipment Corporation.
Over the previous year the PDP-10 had taken the computing world by storm. One of DECs series of commercially successful units that had grown out of the artificial intelligence programs at MIT and Stanford, the PDP-10 was exceptional: A relatively compact computer (though it still took up a good-sized room) with time-sharing hardware that rapidly replaced the aging SDS 940s in scores of college computer centers and corporate back offices. The PDP-10 was so well designed that it is still treated with a cliquish reverence by computer jocks today, more than a decade after the last unit rolled off the factory line.
The PDP-10 was also becoming the computer of choice—a de facto standard—for resea
rch departments linked to the burgeoning ARPANET. It possessed the special virtue of having been specifically tailored to run Lisp, a popular programming language in which ARPANET researchers turned out software by the ream. Moreover, its operating system, Tenex, had been developed on an ARPA contract with government funds, and was thus available for anyone to use, free of charge.
By the mid-1970s, thanks to these features and others, computer science in the United States would cease to be a Babel of dozens of mutually incompatible machines. Instead, the prototypical computer research group would be a university department using PDP-10s equipped with the Tenex operating system and linked to the ARPANET." Even as early as 1971, it was hard to imagine a better description of CSL's self-image. After all, Taylor was the ARPANET'S patriarch and they were almost all ARPANET "brats," as they labeled themselves. To the Computer Science Lab it was almost an article of faith that PARC should be the PDP-10 node par excellence.
But the lab was about to discover that purchasing capital equipment for a corporate facility could never be as straightforward a question of engineering as it might be for a faculty department at, say, MIT. For
*In 1979, ARPANET officials reported that of 202 host computers connected to the network, 104 were PDP-10s or their corporate relatives, PDP-11s.
when word got out that Bob Taylor was planning to install a DEC computer at PARC, the executives at SDS went off like Roman candles. This was understandable. In 1971 SDS was fighting for its life, and DEC was its principal foe.
Not long after completing its purchase of SDS Xerox had discerned that all was not well at its new computer division. Max Palevsky, who had personally pocketed some $100 million from the billion-dollar transaction, was happier spending his money on Democratic campaigns and movie and magazine ventures ("I must be on every sucker list in town," he said cheerily) than burying himself in the day-to-day problems of his old firm. Suspicion was rife that his real business talent resided in his knowing how to get out while the getting was good; popular rumor even had him confiding to aides about the SDS deal: "We sold them a dead horse before it hit the ground." He denied the epigram, but acknowledged knowing that at the time of the sale SDS had entered a rocky stretch of the business cycle.
As though to confirm everyone's worst suspicions, SDS profits peaked in May 1969, at almost the very moment that ownership formally changed hands. That calendar year SDS recorded about $12 million in net income. The following year was its best ever in terms of sales—but for every dollar it recorded in revenue, it lost two. The division never again showed a profit. Over the following three years the losses would add up to nearly $120 million.
No later than early 1970, Rigdon Currie recalled, "the handwriting was on the wall." At that time Currie, as head of sales, was instructed to prepare a detailed presentation on SDS prospects for a management meeting to be chaired by the choleric ex-West Pointer Dan McGurk, a former Palevsky lieutenant who had become the divisions president following the sale.
Currie was a Georgia Tech-trained engineer whose southern courtliness masked a wicked sense of humor. He was determined to make sure his presentation would not easily be forgotten. On the appointed day he greeted the assembled executives while standing by a flip chart three feet tall. From the outset it was clear the news was not good. The condition of every one of the company's leading customers was dire, he said. Government agencies like NASA and the National Science Foundation were sharply cutting their budgets. Time-sharing companies like Tymshare, which ran its nationwide network on SDS 940s, were encountering sales resistance at all levels. With the education, government, and aerospace industries all slumping simultaneously, nearly every market SDS served was on the rocks.
Moreover, thanks to the PDP-10, SDS no longer found itself on the cutting edge of the time-sharing market, where it had been placed by the work of Project Genie. Even long-term customers turned in their leases on the SDS 940 once they got a look at the new DEC machine. That was an ominous trend. The old SDS leases were a fiscal time bomb: They permitted the customers to return their machines prior to expiration for a full pro-rata credit on the unexpired period. Unfortunately, SDS had long since recorded the full-term lease payments as revenue. This meant that as computers came back from failed or defecting clients SDS would have to recalculate the profits it had already reported for earlier quarters and quite possibly restate them—as losses.
"Now, taking all these factors together, where are we?" Currie asked out loud. He turned over the top card on his flip chart. The next one was blank except for two words in huge block letters: "DEEP SHIT."
"McGurk almost killed me," he remembered.
Adding to the predicament of shrinking markets was Peter McColough's determination to push SDS into a new one: the business data processing business ruled by IBM.
Although McColough has been roundly chastised for this strategy, it may actually have originated with Max Palevsky. The SDS founder had always weighed his company against IBM as a benchmark. To outsiders this resembled weighing a puppy against a horse; no matter how big it grows, it can never play in the same league. Yet in the weeks following the Xerox sale Palevsky had openly fantasized about placing Xerox's capital base behind the engineering skills of SDS and marching jointly into war against Big Blue. (Much later he hinted he was only telling McColough what the Xerox CEO wanted to hear.)
To the rest of the industry the act of stepping into IBM's den seemed a fool’s errand. Big Blue did not always offer tire most technologically advanced products in the market and frequently charged the highest prices, but the ruthlessness of its sales and marketing more than made up for these shortcomings (and fomented an epic battle with federal antitrust authorities). Companies that tried to dislodge IBM from its perch generally got butchered in the attempt. That had been the fate even of the industrial powerhouses General Electric and RCA, which had given up after losing a combined $600 million in ten years of trying.
The only major competitors left in the market were five large computer makers known familiarly as the "Bunch." The term was an acronym of their names—Burroughs, (Sperry) Univac, NCR, Control Data, and Honeywell—but it was also descriptive. The Bunch trailed IBM in a distant cluster, like a field of thoroughbreds struggling to keep up with Secretariat. From 1960 through 1970 the five competitors lost a combined $167 million while IBM racked up profits of more than $3.5 billion.
Into this valley of death Peter McColough now proposed to ride SDS. The very notion all but proclaimed his ignorance of the computer industry in general and his own subsidiary in particular, for there could scarcely exist a company less qualified to take on this particular fight.
"We did not have the tools to do it," Robert Spinrad recalled.
Bob Spinrad was an experienced scientist and engineer, unassertive but rigorously analytical. He was a gregarious gentleman of the old school, with a round face marked by eyebrows as dark and emphatic as exclamation points. Spinrad had joined SDS in 1968 to run its software group after several years at Brookhaven National Laboratory, where the computer of choice was an SDS machine. Now he found himself swept unwillingly into McColough’s foolhardy crusade. Shortly after the chief executive s decree, Spinrad met with a couple of Xerox officials sent belatedly from Stamford to El Segundo to appraise the division's competitive ability in the new market.
"How good is your COBOL compiler?" they asked. COBOL was the tedious programming language used for repetitive and uncomplicated business programs such as payrolls and budgets. On hearing the question, Bob Spinrad recognized as though for the first time the enormity of the task confronting the company. Scientific and research programmers, like those who worked for SDS and its traditional customers, would not be caught dead working in COBOL, which they considered a lame language suitable only for clerks and drones. He shifted uneasily in his chair.
"It's not a question of how good our COBOL compiler is," he told the visitors.
"Why not?"
&
nbsp; "Because we don't have one."
Xerox had thoroughly misunderstood the difference between scientific computing, in which SDS might with great effort manage to hold its own, and business computing, in which it was a non-starter. The company could not offer a fraction of the product line any business client would expect as a matter of course, whether high-speed line printers or robust database programs. As Spinrad recalled, "A lot of things that were ho- hum standard operating procedure for companies that served a commercial environment just weren't in our lexicon or armament."
The harvest was a profound morale crisis in El Segundo, where every employee understood what the company's limitations were, how hard it would be to change course, and how determined Xerox was to do it anyway.
One weekend McGurk summoned his department heads to a weekend "retreat" in a hotel near Los Angeles International Airport to iron out the difficulties. They broke into study groups, turned over the issues late into the night, and reconvened around a very long conference table on Sunday afternoon.
Dealers of Lightning Page 13