conanxin

Compiled from: An article introducing PARC in the 1985 IEEE Spectrum - Inside the PARC: the `information architects'

At the end of 1969, Xerox Chairman C. Peter McColough told the New York Society of Security Analysts that Xerox was determined to develop "information architecture" to address the problems brought about by the "knowledge explosion." Legend has it that McColough then turned to Jack E. Goldman, the Senior Vice President of Research and Development, and said, "Well, go establish a lab and find out what I just meant."

Goldman's account differs. In 1969, Xerox had just acquired a large computer manufacturer, Scientific Data Systems (SDS). "When Xerox acquired SDS," he recalled, "I quickly walked into Peter McColough's office and said, 'Look, since we're in the digital computer business, we better have a research lab!'"

In any case, the result was the establishment of the Xerox Palo Alto Research Center (PARC) in California, one of the most unusual corporate research institutions of our time. PARC is one of three research centers at Xerox; the other two are in Webster, New York, and Toronto, Canada. It employed about 350 researchers, managers, and support staff (in contrast, AT&T's Bell Labs employed about 25,000 before its breakup). PARC had been established for 15 years and had produced or nurtured several technologies that led to developments including:

• Macintosh computers with a mouse and overlapping windows.
• Colorful weather maps in television news programs.
• Laser printers.
• VLSI system design structures, now taught in over 100 universities.
• Networks connecting personal computers in offices.
• Semiconductor lasers for reading and writing optical disks.
• Structured programming languages like Modula-2 and Ada.

By the mid-1970s, nearly half of the top 100 computer scientists in the world were working at PARC, which had similar strengths in other fields, including solid-state physics and optics.

Some researchers say that PARC was a product of the civil rights philosophy of the 1960s and that decade's focus on improving the quality of life. The center was established in 1970, unlike other major industrial research labs; its work was not tied to the product lines of its corporate parent. Unlike university research labs, PARC had a unified vision: it would be dedicated to developing "information architecture."

The origin of this phrase is unclear. McColough credited it to his speechwriter. The speechwriter later said that neither he nor McColough had a specific definition for the phrase.

Thus, almost everyone who joined PARC at its inception had a different view of the center's charter. This had its advantages. Since projects were not assigned from above, researchers formed their own small groups; support for a project depended on how many people the project's planner could get involved.

"The phrase is 'Tom Sawyer,'" recalled James G. Mitchell, who joined PARC in 1971 from the now-defunct Berkeley Computer Company and is now Vice President of Research at the Palo Alto Acorn Research Center. "Someone would think something was really important. They would start working on it, form a group, and then try to persuade others to join them."

The First Step#

When Goldman established PARC, his first decision was to bring in his old friend George E. Pake to manage it. Pake was the Executive Vice President, Provost, and Professor of Physics at Washington University in St. Louis, Missouri. Pake's first decision was to hire Robert Taylor, who was then at the University of Utah, to help him recruit engineers and scientists for the Computer Science and Systems Science Laboratory.

Taylor had served as the Director of the Information Processing Techniques Office at ARPA (Advanced Research Projects Agency) and, in the mid-1960s, he and others funded the heyday of computer research.

PARC started as a small organization—possibly fewer than 20 people. Nine of them came from the Berkeley Computer Company, a small mainframe company that Taylor had tried to persuade Xerox to acquire as a way to kickstart PARC. (Many at BCC were responsible for the design of the SDS 940, which Xerox acquired in 1968.)

The 20 PARC employees lived in a rented small building, "using rented chairs, rented desks, with a four-button telephone, and no receptionist," recalled David Thornburg, who joined PARC's Integrated Sciences Laboratory in 1971 after graduating from graduate school. The organization felt it should have its own computer.

Mitchell said, "It was a bit difficult to do language research and compiler research without machines." The computer they wanted was the PDP-10 from Digital Equipment Corporation (DEC).

Alan Kay recalled, "There was competition between Xerox's SDS and DEC in an ad in Datamation [magazine]." At the end of 1970, he joined PARC as a researcher from Stanford University's Artificial Intelligence Lab. "When we wanted a PDP-10, Xerox envisioned a photographer in the PARC lab capturing a DEC box, so they said, 'How about a Sigma 7?'"

"We thought it would take three years to get an operating system for the Sigma 7, while we could build a complete PDP-10 in a year."

The result was the MAXC (Multiple Access Xerox Computer), which simulated the PDP-10 but used semiconductor dynamic RAMs instead of a core. MAXC invested heavily in hardware and software and maintained a historical record of continuous availability as a node on ARPAnet.

MAXC was crucial to many developments. Intel produced the 1103 dynamic memory chips used in the MAXC design, gaining initial profits. Kay recalled, "Most of the 1103 memory chips purchased from Intel at the time didn't work." Therefore, PARC researcher Chuck Thacker created a chip tester to screen chips for the MAXC. Later versions of the tester, based on the Alto personal computer, were also developed at PARC and eventually used by Intel in its production line.

Moreover, MAXC provided PARC with experience in building computers, which was beneficial for the center. "If we had bought a PDP-10, we wouldn't have acquired the three capabilities we needed," recalled an early lab manager at PARC. "We needed to develop a vendor community—to complete design layouts, printed circuit boards, and so on—the only way to drive this was with a project. We also needed semiconductor memory, which PDP-IOs didn't have. We thought we needed to learn more about programmable microprogrammed machines, even though we didn't use those features."

MAXC set a pattern for PARC: building its own hardware. This gave researchers a vision that had to become reality—at least on a small scale.

Kay said, "The original founders swore we would never build a system that wasn't designed for 100 users. This meant that if it was a time-sharing system, you had to run 100 people on it; if it was a programming language, then 100 people had to program without having to hold on to it all the time. If it was a personal computer, we had to be able to build 100 of them."

This policy of building systems was not the only research approach; Mitchell recalled it was a focal point of debate at PARC.

"System research requires building systems," he said. "Otherwise, you don't know if your ideas are good or how difficult they are to implement. But some people think that when you're building something, you're not doing research."

Since building the MAXC, the center had produced prototypes of dozens of hardware and software systems—sometimes thousands of prototypes.

The first personal computer developed in the United States is often considered to be the MITS Altair, which was sold in 1976 as a kit for business enthusiasts. Almost at the same time, the Apple I was also available, similarly in kit form.

However, by the end of that year, there were also 200 Alto personal computers in everyday use, the first of which was built in 1973. While researchers in PARC's Computer Science Laboratory were finishing the MAXC and starting to use it, their peers in the Systems Science Laboratory were assembling a distributed computing system using Nova 800 processors and high-speed character generators. In September 1972, PARC Computer Science Laboratory researchers Butler Lampson and Chuck Thacker approached Alan Kay from the Systems Science Laboratory and asked, "Do you have funding?"

Kay told them he had about $250,000 earmarked for purchasing more Nova 800 and character generation hardware.

"Do you want us to build you a computer?" Lampson asked Kay.

"I would love to," Kay replied. On November 22, 1972, Thacker and Ed McCreight began building the Alto. A Xerox executive reportedly insisted that developing a large hardware system would take 18 months, which infuriated Thacker. When Thacker argued that he could do it in three months, he made a bet.

It took just over three months, but not much beyond that. On April 1, 1973, Thornburg recalled, "I walked into the basement where the prototype Alto was located, with wires connected to a rack full of Novas, and saw Ed McCreight sitting in a chair with 'Alto lives' written in the upper left corner of the display."

Kay said the Alto proved to be "what Lampson needed, what Thacker needed, and what I needed. Lampson wanted a $500 PDP-10," he recalled. "Thacker wanted a machine that was ten times faster than the Nova 800, and I wanted a machine that could be carried around and used by children."

The rapid construction of the Alto was due to its simplicity. Kay recalled that the processor "was just a timer"—with only 160 chips in the original integrated circuit technology of 1973. This architecture traced back to the TX-2, built by MIT's Lincoln Laboratory in the late 1950s with 32 program counters. The Alto, with 16 program counters, would fetch its next instruction from the highest-priority counter at any given time. Executing multiple tasks incurred no additional overhead. When the machine drew the screen display, dynamic memory refreshed every 2 milliseconds, the keyboard was monitored, and information was transferred in and out from the disk. The lowest-priority task was running the user's program.

The prototype was successful, and more Altos were built. Serious research began on user interfaces, computer languages, and graphics. Lampson, Thacker, and other project planners received the first models. Many PARC researchers pitched in to speed up production, but it seemed that demand was never fully satisfied.

"There was a lab producing Altos, surrounded by circuit boards, and anyone could come in and work," recalled Daniel H.H. Ingalls, who is now Chief Engineer at Apple Computer in Cupertino, California.

Ron Rider, who was still at Xerox, "had an Alto when they were unavailable," Bert Sutherland recalled, who joined PARC in 1975 as the manager of the Systems Science Laboratory. "When I asked him how he got it, he told me he went around to various labs, collected parts people gave him, and then assembled it himself."

Networking#

By today's standards, the Alto was not a particularly powerful computer. However, if several Altos were connected to file servers and printers, the result looked like the office of the future.

Before PARC was established—in 1966, at Stanford University—the idea of local computer networks had already been discussed. Larry Tesler, now a manager of object-oriented systems at Apple, graduated from Stanford when the university was considering purchasing IBM's time-sharing system.

"I suggested to one of them that they buy 100 PDP-1s and connect them to a network," Tesler said. "Some consultants thought it was a good idea; a consultant from Yale, Alan Perlis, told them it was the way to go, but the IBM-oriented people at Stanford thought it would be safer to buy a time-sharing system. They missed the opportunity to invent local networking."

So PARC ultimately gained another first. While building the Alto, Thacker envisioned Ethernet, a coaxial cable that connected machines in the simplest way possible. It was partly based on a packet radio network called Alohanet developed at the University of Hawaii in the late 1960s.

Kay said, "Thacker said coaxial cable was just capturing ether, so that part was determined before Robert Metcalfe and David Boggs showed up—this would be packet switching, and it would be a collision network. But Metcalfe and Boggs spent a year figuring out how to do the damn thing. (Metcalfe later founded 3Com in Mountain View, California; Boggs now works at DEC Western Research in Los Altos, California. Both hold the basic patents for Ethernet.)

"I always thought it was important that Boggs was an amateur radio operator," Sutherland said. "That had a significant impact on how Ethernet was designed because Ethernet fundamentally could not work reliably. It was like citizen band radio or any other type of radio communication; fundamentally, it was unreliable, just like our view of the telephone. Because you knew it basically didn't work, you did all the fault-tolerant programming—'Say it again, it's garbled' protocols designed for radio communication. This made the final network function very reliably."

"Boggs was an amateur radio enthusiast who knew you could communicate reliably over unreliable media. I often wonder what would have happened if he didn't have that background," Sutherland added.

Once Ethernet was built, it was relatively simple to use: a computer wanting to send a message would wait and check if the cable was clear. If it was clear, the machine would send the information in a data packet beginning with the recipient's address. If two messages collided, the machines sending them would each wait a random amount of time before trying again.

One innovative use of the network had nothing to do with people sending messages to each other; it only involved communication between machines. Because dynamic memory chips were so unreliable in those days, when nothing else was happening, the Alto would also perform memory checks. Thornburg said it was very significant in detecting bad chips: "It would send a message telling you which Alto was bad, which slot had a bad board, and which row and column had bad chips. I found this out because one day a repairman came over and told me, 'Anytime you're ready to shut down, I need to fix your Alto,' and I didn't even know what was wrong."

While developing Ethernet, another key factor for the future office was the laser printer. After all, what good is a network that can display documents in multiple styles if there is no effective means to print them and transfer documents from one place to another?

The idea for the laser printer came from Xerox's Webster Research Lab in New York, championed by Gary Starkweather. At the time, then Vice President of Research Goldman recalled that his idea was to use a laser to draw information in digital form onto the photocopier's selenium drum or belt. Starkweather reported to George White, Vice President in charge of the Advanced Development Business Products Group.

"George White came to me," Goldman said, "Listen, Jack, I found an amazing guy named Gary Starkweather who prints visual information with lasers, of course using Xerox machines. What an ideal concept for Xerox. But I don't think he will get anywhere in Rochester; no one will listen to him, and they won't do anything very leading-edge. Why don't you bring him to the new lab in Palo Alto?"

The newly appointed PARC manager Pake seized the opportunity. Starkweather and several other researchers from Rochester were transferred to Palo Alto and started PARC's Optical Sciences Laboratory. The first laser printer, EARS (Ethernet-Alto-Research character generator-Scanning laser output terminal), built by Starkweather and Ron Rider, began printing documents generated by Altos and sent them over Ethernet in 1973.

Thornburg said EARS was not perfect. It had a dynamic character generator that could create new patterns for characters and graphics as they came in. If there was no capital "Q" on the page, the character generator would save internal memory by not generating a pattern for the capital "Q." But if the page contained a very complex image, the character generator would run out of patterns; "There was a certain degree of complexity in the unprintable drawings," Thornburg recalled.

Despite these shortcomings, the laser printer was a significant advancement over the line printers, teletypes, and fax machines used at the time, and Goldman pushed for its commercialization as soon as possible. However, Xerox refused. In fact, one sore point in PARC's history was that the parent company seemed unable to capitalize on the developments of its researchers.

In 1972, when Starkweather built the first prototype, Lawrence Livermore National Laboratory proposed a bid for five laser printers to promote the technology. However, Goldman could not persuade the executive at Xerox's Electro-Optical Systems Division (whose background was in accounting and finance) to allow the bid. The reason was: if the laser printer required repairs as frequently as a copier, Xerox could lose $150,000 during the contract period, even though initial evidence suggested that wear and tear from printing was much less than from copying.

In 1974, when a small group of PARC researchers led by John Ellenby began purchasing used copiers from Xerox's copier division and installing laser heads in them, the laser printer finally emerged outside PARC. John Ellenby built the Alto II, a production version of the Alto, and he is now Vice President at Grid Systems in Mountain View, California. The resulting printers were called Dovers and were used internally at Xerox and in universities. Sutherland estimated that dozens of such printers were manufactured.

He recalled, "They took all the optical instruments out of the printers and sent them back to the copier division." He said that even today, he still receives laser-printed documents from universities where he can recognize the Dover font.

Also in 1974, the product review committee at Xerox's headquarters in Rochester finally made a decision on what kind of computer printer the company should produce. Goldman said, "A group of people who knew nothing about technology was making decisions, and a week before the decision was made, I saw it heading toward CRT technology." (Another group at Xerox developed a printing system that focused text displayed on a special cathode ray tube onto the copier's selenium drum and printed it.)

"That was Monday night. I commandeered a plane," Goldman recalled. "I told the planning vice president and the marketing vice president, 'You two come with me. Adjust your plans for Tuesday. Tonight, you're going with me to PARC. We'll come back for the 8:30 AM meeting on Wednesday.' We left around 7 PM and arrived in California at 1 AM, and God bless, the PARC folks did a beautiful demonstration showing what the laser printer could do.

"If you're dealing with marketing or planning people, let them experience it firsthand. All the charts and slides are useless," Goldman said.

The product review committee chose laser technology, but there were delays. "They wouldn't let us bring them out in the 7000 series," Goldman said, referring to the old printers used by Ellenby's team. "Instead, they insisted on launching a new 9000 series, which didn't come out until 1977."

From a purely economic perspective, Xerox's return on investment in PARC's first decade came from the profits of the laser printer. This was perhaps ironic, as one vision of the future office was paperless.

"I think PARC produced more paper than any other office because with the press of a button, you could print 30 copies of any report," noted Douglas Fairbairn, a former PARC technician and now Vice President of User Design Technology at VLSI Technology. "If the report was 30 pages long, that's 1,000 pages, but it still only took a few minutes. Then you say, 'I want that picture on another page,' and that's another 1,000 pages."

By the mid-1970s, most Altos in PARC researchers' offices could be customized according to their needs. Richard Shoup's Alto had a color monitor. Taylor's Alto had a speaker that played "The Eyes of Texas Are Upon You" whenever he received an email.

Moreover, in the ten years since the Alto was widely used at PARC, it became evident that personal computers could be used for both entertainment and work. PARC researchers were among the first to discover this.

"At night, whenever I was in Palo Alto," Goldman said, "I would go to the lab to see Alan Kay creating a game. This was before video games appeared; these kids were creating these things until midnight or 1 AM."

Sutherland said, "I enjoyed observing some firsts; Xerox held the first electronic lottery nationwide. At Xerox, I received my first electronic junk mail, my first electronic work acceptance, and my first electronic obituary."

When the Xerox 914 copier was introduced in the early 1960s, "I was a copying maniac," Lynn Conway said, who joined PARC in 1973 from Memorex and is now the Vice Dean and Professor of Electrical Engineering and Computer Science at the University of Michigan. "I loved making things and distributing them, like maps—various things. In the Xerox environment of 1976, suddenly you could create a lot of things."

Dozens of clubs and interest groups formed on the network. No matter what PARC employees' hobbies or interests were, they could find someone to share their interests electronically. Many serious works were also completed electronically: reports, articles, and sometimes entire design projects were done through the network.

One side effect of all this electronic communication was the neglect of appearances and other external identities.

John Warnock said, "PARC people tended to have very strong personalities, and sometimes at design meetings, those personalities were stronger than the technical content." He joined PARC in 1978 from Evans & Sutherland, where he worked on high-speed graphics systems. Working through email eliminated personality issues during the design phase. Electronic interaction was particularly useful for software researchers, who could send code back and forth.

Warnock, now President of Adobe Systems in Palo Alto, described the design of Interpress, a printing protocol: "One designer was in Pittsburgh, one in Philadelphia, we had three in the area, and a couple in El Segundo, California. The design was almost entirely done remotely through the mail system; only twice did we all gather in the same room."

Email was also crucial for tracking team projects.

Warren Teitelman, who joined PARC from BBN in 1972, said, "One of the really useful capabilities was to save a series of information about a specific topic so you could refer back to it." He is currently the programming environment manager at Sun Microsystems in Mountain View. Teitelman added, "Or if someone came in late and didn't understand the background, you could bring them up to speed by sending them all the information."

But email sometimes got out of control at PARC. Once, after a week without contact, Teitelman logged into the system and found 600 emails in his inbox.

Superpaint#

Anyone who has attended a business meeting knows that today's offices include graphics and text. In 1970, Shoup, now Chairman of Aurora Systems, began researching new ways to create and manipulate images digitally in the future office at PARC. His research pioneered the field of television graphics and earned him and Xerox an Emmy Award.

Shoup recalled, "It quickly became clear that if we wanted to do raster scanning systems, we should do it compatible with television standards so we could easily obtain monitors, cameras, and VCRs." In early 1972, he built some simple hardware to generate anti-aliased lines, and by early 1973, the system named Superpaint was completed.

AIvy Ray Smith recalled that this was the world's first complete drawing system with an 8-bit frame buffer; he worked on Superpaint at PARC and would soon become Vice President and Chief Technology Officer at Pixar in St. Raphael, California. It was also the first system to use multiple graphic assist tools: a color lookup table for simple animations, a digitizing tablet for input, and a palette for mixing colors directly on the screen. The system also had a real-time video scanner, allowing images of real objects to be digitized and manipulated.

Shoup said, "The first thing I did on this system was some anti-aliased lines and circles because I wrote a paper on the subject but didn't finish those examples. But when I submitted the paper and it was accepted, the machine used to make the examples hadn't been built yet."

By mid-1974, additional software enhanced Superpaint, allowing it to perform a variety of tricks. Smith had just completed his doctoral work in a mathematical branch called cellular automata theory and was hired to help test the machine. He made a tape called "Vidbits" with Superpaint, which was later exhibited at the Museum of Modern Art in New York.

Six months later, his initial contract with PARC expired, and he did not renew it. Although Smith was disappointed, he was not surprised, as he found that not everyone enjoyed painting with computers.

"The color graphics lab was a long narrow room with seven doors leading into it," he recalled. "You had to go through it to get to many other places. Most people would walk by, look at the screen, and stop—even the most mundane things they had never seen before. Bicycle color maps had never been seen before. But some people would walk by without stopping. I can't imagine how people could walk through that room without stopping to take a look."

In addition to others' indifference to video images, another reason for Smith's departure might have been the public television program where many viewers saw Superpaint for the first time, a show called "Supervisions" produced by Los Angeles KCET. "It only took a few uses to produce very few color cycling effects," Shoup recalled. However, Xerox was not pleased with the unauthorized use of the system on the show. "Bob Taylor sat with Alvy [Smith] for an entire afternoon, and Alvy pressed the erase button on the VCR, removing Xerox's logo from every copy of the tape," Shoup continued. (This was one of the tapes the committee saw when awarding Xerox the Emmy.)

Shoup remained at PARC, supported by Kay's research group, while Smith moved on, receiving funding from the National Endowment for the Arts to work on computer art. He received support from the New York Institute of Technology, where he helped develop Paint, which became the basis for Ampex Video Art (AVA) and N.Y. Tech’s Images, which still use the system today.

While Shoup was working on Superpaint alone at PARC, he was not the only Superpaint enthusiast nationwide searching for frame buffers. David Miller, now known as David Em, and David DiFrancesco were among the first artists to paint with pixels. When Em lost access to Superpaint, he began a year-long search for frame buffers, eventually entering the Jet Propulsion Laboratory in California.

Finally, in 1979, Shoup left PARC to start his own company to manufacture and sell paint systems—Aurora 100. He admitted that he did not achieve any technical leaps in designing Aurora; it was merely a second-generation version of his first-generation system at PARC.

Shoup said, "The Aurora-based machines we made for the next generation were directly related to what we were thinking about seven or eight years ago at PARC."

Aurora 100 is now used by companies for internal training films and demonstration graphics. Today, thousands of artists are painting with pixels. At Siggraph's art show in San Francisco in 1985, 4,000 entries were received.

Mouse and Modes#

Most people who know that the mouse is a computer peripheral think it was invented by Apple. Experts would correct them, saying it was developed at Xerox PARC.

But in fact, the mouse appeared before PARC. "I saw a demonstration of the mouse being used as a pointing device in 1966," Tesler recalled. "Doug Engelbart at the Stanford Research Institute invented it."

At PARC, Tesler began proving that the mouse was a bad idea. "I really didn't believe in it," he said. "I thought the cursor keys were better."

"We tested some people who had never seen a computer. Within three or four minutes, they happily edited using the cursor keys. At that point, I was preparing to show them the mouse and prove that they could select text faster with the mouse than with the cursor keys. Then I was going to prove they didn't like it."

"Contrary to expectations. I would let them spend an hour using the cursor keys, which made them really accustomed to those keys. Then I would teach them how to use the mouse. They would say, 'This is fun, but I don't think I need it.' Then they would play with the mouse for a while, and two minutes later, they would never touch the cursor keys again."

After Tesler's experiments, most PARC researchers accepted the mouse as an appropriate peripheral for the Alto. One person who did not like the mouse was Thornburg.

"I didn't like the mouse," he said. "It was the least reliable component of the Alto. I remember going to the PARC repair room—there was a shoebox for good mice and a 50-gallon drum for bad mice. And it was expensive—too expensive for the mass market."

"While I didn't mind using the mouse for text manipulation, I thought it was completely unsuitable for drawing. In the Stone Age, people stopped painting with stones for a reason: stones are not suitable painting tools; people turned to using sticks."

Thornburg, a metallurgist who had been doing materials research at PARC, began exploring alternative pointing devices. In 1977, he invented a touch tablet and connected it to an Alto. Most people who saw it said, "This is nice, but it's not a mouse," Thornburg recalled. His touch tablet eventually became a product: Koalapad, a home computer peripheral priced under $100.

"It was clear that Xerox didn't want to do anything with it," Thornburg said. "They didn't even apply for patent protection, so I told them I liked it. After a lot of pitching, they said OK."

Thornburg left Xerox in 1981, worked at Atari for a while, and then co-founded a company with another former PARC employee—now Koala Technologies—to manufacture and sell Koalapad.

Meanwhile, although Tesler accepted the mouse as a pointing device, he was not satisfied with how SRI's mouse worked. "The left hand had a five-key key group, and the right hand had a three-button mouse. You would tap one or two keys with your left hand, then point to something with your right hand using the mouse, and then there were more buttons on the mouse to confirm your commands. A command required six to eight keystrokes, but you could operate both hands simultaneously. Experts could complete tasks very quickly."

The SRI system's mode was very complex. In a modal system, the user first indicates what they want to do—for example, a delete operation. This would put the system in delete mode. Then the computer waits for the user to indicate what they want to delete. If the user changes their mind and tries to do something else, they cannot do so unless they first cancel the delete command.

In a non-modal system, the user first points to the part of the display they want to change and then indicates what should be done to it. They could point at things all day, constantly changing their mind, and never have to execute a command.

To complicate matters for the average user (but more efficiently for programmers), the meaning of each key varied depending on the mode the system was in. For example, "J" meant scroll, and "I" meant insert. If the user tried to "insert" and then "scroll" without canceling the first command, they would end up inserting the letter "J" into the text.

Most PARC programmers liked the SRI system and began adapting it in their projects. "Many people thought it was the perfect user interface," Tesler said. "Whenever someone suggested changing it, they were met with hostile stares." As programmers, they had no objection to the fact that the keyboard responded to combinations of keys pressed simultaneously, which were represented in binary symbols for the alphabet.

Tesler began testing the interface with non-programmers. He taught a newly hired secretary how to operate the machine and observed her learning process. "Clearly, no one had ever done this before," he said. "She had a lot of problems with the mouse and keys."

Tesler advocated for a simpler user interface. "The only person who agreed with me was Alan Kay," he said. Kay supported Tesler's attempt to write a non-modal text editor on the Alto.

Although most popular computers today use non-modal software, with the Macintosh being perhaps the best example, Tesler's experiments did not resolve the issue.

"MacWrite, Microsoft Word, and Xerox Star all started as projects with complex modes," Tesler said, "because programmers didn't believe a user interface could be flexible, useful, and extensible unless it had many modes. It turned out that this was not achieved through persuasion; customers complained that they liked extremely simple non-modal editors, with no better functionality than this editor, because this editor had all the features they couldn't think of how to use."

Kids and Us#

Similarly simplified non-modal editors also applied to PARC's programming languages and environments. In search of a language that children could use, Kay could often be seen testing his work among kindergarten and elementary school students.

Kay's goal was the Dynabook: a simple, portable personal computer that could meet a person's information needs and provide a channel for creativity—writing, drawing, and music creation. Smalltalk was the language of the Dynabook. It was based on the concept of classes advocated in the programming language Simula and the idea of interactive objects that communicate by sending messages requesting actions rather than directly manipulating data.

The first version of Smalltalk resulted from a chance conversation between Kay, Ingalls, and another PARC researcher, Ted Kaehler. Ingalls and Kaehler were considering writing a language when Kay said, "You can write one on a page."

He explained, "If you look at the Lisp interpreter itself, the kernel of these things is very small. The kernel of Smalltalk might even be smaller than Lisp."

Kay recalled that the problem with this approach was that "Smalltalk is doubly recursive: before you do anything to the parameters, you're already using functions." In the first version of the language, Smalltalk-72, controls were passed to objects as quickly as possible. Thus, writing a concise Smalltalk definition in Smalltalk was very difficult.

"Writing ten lines of code took about two weeks," Kay said, "and it was hard to tell if those ten lines of code were valid." Kay spent two weeks thinking from 4:00 AM to 8:00 AM, then discussing his ideas with Ingalls. When Kay finished, Ingalls wrote the first Smalltalk in Basic on the Nova 800, as that was the only available language with good debugging capabilities at the time.

Because the language was very simple, the speed of developing programs and even entire systems was quite fast. Kay said, "Smalltalk was so large that you could go out for a beer or two and come back, and two people would inspire each other to complete a full system in an afternoon." From one afternoon's development, overlapping windows emerged.

The concept of windows originated from Sketchpad, an interactive graphics program developed by Ivan Sutherland at MIT in the early 1960s; Evans & Sutherland implemented multiple windows on a graphics machine in the mid-1960s. However, PARC's Diana Merry implemented the first multiple overlapping windows on the Alto in 1973.

"We all thought the Alto display was very small," Kay said, "and it was clear that without a large display, you had to have overlapping windows."

Following the windows came the concept of Bitblt—transferring blocks of data from one part of memory to another without restrictions on word boundaries. Thacker, the main designer of the Alto, implemented a function called CharacterOp that wrote characters to the Alto's bitmap screen, and Ingalls extended that function into a general-purpose graphics tool. Bitblt made overlapping windows simpler and also enabled various graphics and animation techniques.

Ingalls recalled, "In early 1975, I did a demonstration for all of PARC's Smalltalk systems, using Bitblt to create menus and overlapping windows and things. Later, a group of people found me and said, 'How did you do that? Can I get the Bitblt code?' Within two months, those things were in use throughout PARC."

Despite being flashy and impressive, Smalltalk-72 "was a dead end," Tesler said, "it was ambiguous. You could read a piece of code and not distinguish between nouns and verbs. You couldn't finish quickly, and you couldn't compile."

The first compiled version of Smalltalk, written in 1976, marked the end of the emphasis on a language that children could use. Ingalls said the language was now "a mature programming environment, and we were interested in outputting it and making it widely known."

The next major version of Smalltalk was Smalltalk-80. Kay no longer argued that any language should be simple enough for children to use. Tesler said Smalltalk-80 went too far in the opposite direction from the earliest versions of Smalltalk: "It became so extreme in being editable, unified, and readable that it actually became difficult to read, and you certainly wouldn't want to teach it to children."

Kay looked at Smalltalk-80 and said, "It is very bad that it cannot be used by children because that was the goal of Smalltalk. It reverted to data-structured programming instead of simulation-based programming."

While Kay's group was developing a language for children of all ages, a group of artificial intelligence researchers at PARC was improving Lisp. Lisp was brought to PARC by Warren Teitelman and Daniel G. Bobrow, who came from Bolt, Beranek, & Newman in Cambridge, Massachusetts, where Lisp was developed as a service for the ARPA community. At PARC, it was renamed Interlisp, adding a window system called VLISP and developing a powerful set of tools for programmers.

In PARC's Computer Science Laboratory, researchers were developing a powerful system programming language. After several iterations, the language became Mesa—a modular language that allowed multiple programmers to work simultaneously on a large project. The key was the concept of interfaces—what modules in a program do, not how they work. Each programmer knew what other modules were authorized to do and could call them to perform specific functions.

Another major feature was Mesa's powerful type-checking capabilities, which prevented programmers from using integer variables where real numbers were needed or using real numbers where strings were needed—and prevented bugs from propagating from one module of a program to another.

These concepts later became widely used as the foundation for modular programming languages. "Many ideas from PARC's programming language research influenced Ada [the U.S. Department of Defense's standard programming language] and Modula-2," said Chuck Geschke, now Executive Vice President at Adobe Systems. In fact, Modula-2 was written by computer scientist Niklaus Wirth after a sabbatical at PARC.

No One is Perfect#

Although the successes of PARC's research may have outweighed its achievements, like any organization, it could not escape some failures. The example most frequently cited by former PARC researchers is Polos.

Polos was an alternative approach to distributed computing. While Thacker and McCreight were designing the Alto, another team at PARC was working on Data General Novas in groups of 12, trying to distribute functions among the machines so that one machine could handle editing, another could handle input and output, and another could handle file archiving.

Sutherland said, "With Altos, everything everyone needed was on each machine. Polos tried to achieve this differently—by functionally partitioning."

When Polos began working, the Alto computers were being rolled out throughout PARC, so Polos was shut down. However, it had a second life: Sutherland distributed 12 Novas in other departments at Xerox, which became the first remote gateways on the PARC Alto network, and Polos displays were used as terminals at PARC until they were retired in 1977.

Another significant PARC failure project was the optical character reader and fax machine combination. The idea was to develop a system that could print pages mixing text and graphics, recognize the text itself, and transmit characters in ASCII code, then send the rest of the content using a less efficient fax encoding method.

Charles Simonyi, an application development manager at Microsoft, said, "This was very complex, quite crazy. In this project, they had this incredible hardware, equivalent to a 10,000-line Fortran program." Unfortunately, at that time, that meant thousands of independent integrated circuits.

Conway, who worked on the OCR project, said, "While we made substantial progress in algorithms and architecture, it was clear that the circuit technology of the time was not economically feasible." The project was canceled in 1975.

Turning Point#

Essentially, PARC researchers worked in an ivory tower for the first five years; while projects were still in their infancy, there was little time to do anything else. However, by 1976, with an Alto on every desk and email as a way of life at the center, researchers were eager to see their work used by friends and neighbors.

Kay recalled that at that time, PARC and other departments at Xerox were using about 200 Altos; PARC suggested that Xerox launch a production version of the Alto: Alto III.

Kay said, "On August 18, 1976, Xerox rejected Alto III."

As a result, researchers did not hand over their projects to the manufacturing department but continued to work with the Alto.

"This is why we failed," Kay said. "We didn't throw away the Altos. Xerox management had long been told that PARC's Altos would be like Kleenex, gone in three years, and we needed a new set that was ten times faster. But when that decisive moment came, there was no capital."

"We held a meeting in Pajaro Dunes, California, called 'Let's burn our disk packs.' We could feel that the second derivative of progress was negative for us," Kay said. "I really should have thrown away everyone's disks."

PARC employees did not embark on entirely new research directions but focused on bringing the results of their past research projects to market as products.

Every few years, Xerox would gather all its managers from around the world to discuss the company's direction. At a meeting in Boca Raton, Florida, in 1977, PARC researchers showcased the systems they had built.

The PARC staff assigned to the Boca Raton demonstration poured their hearts, souls, and a lot of Xerox's money into the effort. They designed and built sets, rehearsed in a Hollywood studio, and transported Altos and Dovers back and forth between Hollywood and Palo Alto. Holding the exhibition in the Boca Raton auditorium took an entire day, and a special air-conditioned truck had to be rented from the local airport to keep the machines cool. But for most Xerox employees, it was their first encounter with PARC's "eggheads."

"PARC was a very strange place for the rest of the company," Shoup said. "Not only in California but also nerdy. They were considered strange computer people, with beards, not bathing or wearing shoes, staring at their terminals late into the night, having no connection with anyone else, basically antisocial nerds. Frankly, some of us left that impression, as if we were above the rest of the company."

It was difficult to get other members of Xerox to take PARC researchers and their work seriously.

"The demonstration went very well, the battle was won, but the patient died," Goldman said. Xerox executives not only saw the Alto, Ethernet, and laser printer, but they even showcased a Japanese word processor. "But the company couldn't bring them to market!" Goldman said. (By 1983, the company did launch a Japanese version of the Star computer.)

One reason Xerox struggled to bring PARC's advances to market was that until 1976, no R&D organization had taken research prototypes from PARC and turned them into products.

"At first, the way technology transfer worked was not clear," Teitelman said. "We took a detached view, thinking someone would pick up these technologies. It wasn't until later that this issue received real attention."

Reaching Again#

Even for R&D organizations, getting Xerox executives to accept products was a hard-fought battle. One example is the Notetaker computer, conceived by Adele Goldberg, a researcher from the Smalltalk group, who is now President of the Association for Computing Machinery and still works at PARC. "Poor Adele," Tesler said. "The rest of us got involved and kept redefining the project."

The Notetaker eventually became a battery-powered, 8086-based computer that could fit under an airplane seat. It ran Smalltalk and had a touchscreen designed by Thornburg. "We had a custom display, we had error-correcting memory, and we generally did a lot of custom engineering that we would normally only do for real products," said Fairbairn, the chief hardware designer of Notetaker.

"In my last year at PARC," Tesler said, "I flew around the country with the Notetaker, talking to Xerox executives. It was the first portable computer that ran in airports. Xerox executives made all sorts of promises: we will buy 20,000 units, just talk to this executive in Virginia, then talk to this executive in Connecticut. The company was so decentralized that they never met together. A year later, I was ready to give up."

Xerox may not have been ready to use portable computers, but other companies were. The Osborne I was launched in 1981, about nine months after Adam Osborne reportedly visited PARC, where Notetaker was prominently showcased.

Using Tools#

As some of PARC's pioneers began to feel anxious in the mid-1970s, others were just starting to look for uses for the magical tools of the future office. One of them was Lynn Conway, who developed a new method for designing integrated circuits using the Alto, network, and laser printer, spreading this method to hundreds of engineers at dozens of institutions nationwide.

When Bert Sutherland joined the Systems Science Laboratory as manager in 1975, he brought Caltech professor Carver Mead "into PARC to create some chaos." Mead was a semiconductor design expert who invented the MESFET (Metal-Semiconductor Field-Effect Transistor) in the late 1960s.

Conway recalled that Sutherland was committed to applying computer graphics to integrated circuit layout, so he naturally considered applying advanced personal computers like the Alto to IC design. Attracted to integrated circuit design after the OCR-Fax project's failure, she conceived a streamlined architecture that could only implement racks and equipment racks. But these racks could become chips as long as they could be designed by people who knew what they should do and how to combine them.

"Carver Mead held a week-long integrated circuit design course at PARC," Fairbairn recalled. "Lynn Conway and I were really excited and wanted to do something."

Conway said, "Then a lot of things really came together."

"When Carver and I exchanged ideas about what was happening in the field of computation and devices, he was able to explain some of the basic MOS design methods that had evolved within Intel. We began looking for ways to promote the structures [those designers] had created." Conway explained that she and Mead were not just using computer tools for design but were simplifying design methods and building tools for improved methods.

"From the mid-1975 to mid-1977, things evolved from a series of projects Bert wanted to do into us mastering everything, with examples, and now it was time to write it down."

In less than two years, Mead and Conway developed scalable design rules, repetitive structures, and other concepts now known as structured VLSI design—to the extent that they could teach it in a semester.

Today, over 100 universities teach structured VLSI design, and it has been used to build thousands of different chips. But in the summer of 1977, the Mead-Conway technology was untested—indeed, it was looked down upon. How could they get it accepted?

"The magic of the PARC environment in 1976-1977 was its sense of power; suddenly, you could create a lot of things and make a lot of things. Not just a piece of paper, but an entire book," Conway said.

That is precisely what she and her colleagues did. Conway said, "We just self-published 'Introduction to VLSI Systems,' and if you didn't look closely, you might think it was a completely reasonable, validated thing."

It looked like a book, and Addison-Wesley agreed to publish it as a book. Conway insisted that this would not have happened without the Altos. "Knowledge would have been fragmented and always fuzzy—we couldn't have produced such a pure form, nor could we have done it so quickly."

One of the tools Conway used most in the final stages of the VLSI project was the network: not only the Ethernet within PARC but also ARPAnet, which connected PARC to dozens of research sites nationwide. Conway said, "Looking back, one clear thing is that people couldn't understand we had a powerful intangible weapon. The PARC environment empowered us to overcome those who thought we were crazy or tried to stop us; otherwise, we wouldn't have had the courage to do what we did."

Dragon#

In 1979, three years after Alan Kay wanted to throw away the Altos like tossing "Kleenex," Dorado, a machine ten times more powerful, finally saw the light of day.

"It should have been built by one of the development organizations because they would use it in some of their products," recalled Severn Ornstein, one of Dorado's designers, now the chair of Computer Professionals for Social Responsibility in Palo Alto. "But they decided not to, so if our lab was going to have it, we would have to build it ourselves. We went through a long painful period during which none of us really wanted to do it."

Ornstein said, "At that time, Taylor was managing the lab, and he handled the whole thing very deftly. He never directly twisted anyone's arm; he presided over the work and kept order in the process, but he did make the lab understand that this was something that had to be done. This was indeed a good thing because it was hard to get Dorado to come to life. A lot was lost."

Ornstein recalled that initially, the designers used a new circuit board technology—so-called multiconductor technology, which connected a single wire to a board for connections, which was a wrong start. However, the Dorado board was too complex for multiconductor technology. When the first Dorado ran, many wondered if there would be a second one.

"Butler Lampson's belief was important," Ornstein said. "He was the only one who believed it could be mass-produced."

In fact, even after Dorado was redesigned to use printed circuit boards instead of multiconductor boards and Dorado began mass production, they were still rare. A former PARC manager recalled, "We never had enough budget to fill the entire community with Dorados; they consumed some every year, so by 1984, not everyone had a Dorado."

Those who had used them were envious. "I had my own Dorado," John Warnock said. "Chuck Geschke was a manager; he didn't get one."

"I had an old Alto and a piece of paper," Geschke said.

The arrival of Dorado allowed researchers to take advantage of bitmap displays and all the other advantages of personal computers, making their projects too difficult to operate on the Alto. "We tried to put Lisp on the Alto, and it was a disaster," Teitelman recalled. "When we got Dorado, we spent eight or nine months discussing what we wanted to see in the programming environment, which would combine Mesa, Lisp, and Smalltalk." The result was Cedar, now recognized as one of the best programming environments.

"Cedar incorporated some of Lisp's advantages into Mesa, such as garbage collection and runtime type checking," said Mitchell of Acorn. Garbage collection is a process that recycles memory space that programs no longer use; runtime type checking allows programs to determine the type of their parameters—whether they are integers, strings, or floating-point numbers—and choose the operations they perform on them accordingly.

Interlisp, the language Teitelman nurtured for 15 years, was also ported to Dorado, forming the basis for a research effort that has now evolved into PARC's Intelligent Systems Laboratory.

PARC's Smalltalk group had become accustomed to their Altos and then built another small computer, Notetaker, which encountered some difficulties when processing Dorados.

Ingalls recalled, "In the early days, we got Smalltalk running on an Alto, and I had to take my Alto home, but the direction of development for Xerox machines was contrary to making it easy for people to take machines home. The next machine, Dolphin, was not portable, and Dorado was impossible—it was a dragon."