Chapter One: The Road Behind
Microsoft, a rather new corporation, may not have matured to the position where it understands how it should act with respect to the public interest and the ethics of the marketplace.
--U.S. District Judge Stanley Sporkin
Nathan Myhrvold's pudgy fingers whirred across the keyboard as he peered into his 21-inch computer monitor. Most Microsoft developers would have killed for such a luxury, but Myhrvold was not just some hacker. He was a Ph.D. physicist who had worked on quantum field theories of gravitation with Stephen Hawking at Cambridge. He was also Gates's handpicked technovisionary, chosen to lead Microsoft into the future. Myhrvold had always indulged himself with the newest and best--whether in technology or gourmet cuisine--even before his own company, DSR, was acquired by Microsoft in 1986.
He finished running a spell check through his latest interoffice memo, "Formats and Protocols for Consumer Information." Myhrvold typed blindingly fast. He had early on rejected the old-fashioned QWERTY keyboard developed to keep typewriter keys from sticking if people typed too fast. Instead he used the DVORAK keyboard, designed in 1936, which allowed him to type up to 100 percent faster.
And Myhrvold typed a lot.
The distribution list for this latest opus included the usual suspects: Bill Gates, Mike Maples, Paul Maritz, Charles Simonyi, along with key developers such as Marlin Eller and Murray Sargent. In total the memo would reach forty-four people, and it would change the direction of the company.
Leaving his secretary to disperse his literary output, Myhrvold walked into Eller's office. It was just two doors down from his in Building 9, one of the original structures that looked out over the fountain in the center of Microsoft's campus. Eller was a bright developer and also conceivably one of Microsoft's more stubborn ones. He had been there for years before Myhrvold arrived, and he maintained the healthy skepticism of an old-timer.
"Marlin, is your group fully staffed yet?" Myhrvold asked.
Eller looked up from his 17-inch screen and stopped typing.
"Ah, yeah," Eller said. "Thanks for sending Gordo."
Nathan grinned. Gordon Whitten was another bright old-timer and completely unmanageable, just like Eller. Myhrvold had decided they were the perfect match, hoping, maybe, that they'd cancel each other out.
Myhrvold turned and headed out the door.
"By the way," he said over his shoulder, "take a look at my memo. I've outlined everything your group is doing."
Eller rolled his eyes.
The memos Myhrvold wrote were internal sales pitches, mostly to convince Gates that the technologies Myhrvold talked about were worth funding. The memos were also territory markers, notifying other executives to stay off Myhrvold's turf. "This is what MY group is doing. WE will handle it." Included, also, were implicit warnings about any outside competition.
Myhrvold's memos often caused heated reactions among people at Microsoft. Developers like Eller gnashed their teeth over them because they included marching orders for technologies that were obviously decades into the future. And while some of Myhrvold's ideas were truly visionary, others were just plain bizarre. Myhrvold was a very broad thinker--he read a lot--but he was not necessarily a deep thinker, People who read Myhrvold's memos couldn't always discriminate between what was real and possible, and what was pure sci-fi. Myhrvold himself never appeared to worry much about that distinction, making his memos all the more dangerous because it was so hard to argue against them. How could you prove he was wrong when he was referring to something that might not happen for another ten years?
Eller had no reason to dread this particular memo. The research in his own group was going beautifully, and there was no reason to think Myhrvold's latest diktat would change that.
He left his office and walked down the corridor to the center of Building 9. All the buildings at Microsoft were shaped like an X. Supply rooms, mail rooms, and cafeterias were all in the center of the X. Microsoft had adopted the building design from IBM. This way the company could optimize the number of window offices.
Eller grabbed a Coke from the refrigerator, a freebie Microsoft provided for all its 11,000-plus employees, and headed for the mail room. He set down his drink and grabbed the pile of papers soon to be coming his way. On top was Myhrvold's memo. It looked thick.
Eller walked back to his office. The even thicker stack of papers on his desk seemed to have accreted since he left just a few minutes before. He much preferred the company's internal E-mail system. With E-mail he could just delete random messages and memos instead of filling up his in-basket and then his garbage can with slush. He spent two hours each morning pouring through hundreds of electronic messages, but as soon as 11:00 A.M. rolled around, he just pushed "delete." If the message was important enough, Eller assumed, the sender would try again.
He sat at his desk and glanced through Myhrvold's twelve-page report. The memo described the projects Myhrvold had in place to secure Microsoft's dominant position in the emerging market represented by the convergence of television and computers. Actually, it wasn't exactly Myhrvold's vision. It was a collection of other people's ideas brought together in coherent form. Myhrvold was simply the only one with the time to put the ideas on paper. Besides, he was one of the few Softies with a secretary to copy and distribute memos.
Eller tilted back his ever-present black beret and took a sip of Coke. Even in the summer he wore his trademark chapeau to keep his balding head out of the sun. In winter he wore a beard, in summer his smiling face shined through. He licked his finger and flipped to page four, which detailed the technology that Eller himself was working on.
Eller was in ACT, the advanced consumer technology group, which Myhrvold had recently set up. Gates had decided to make Microsoft the first software company with an internal division fully dedicated to advanced research. It would serve two purposes: to develop add-on products for Windows, and, as analysts have often speculated, to absorb some of the company's outrageously high profits, and thereby, ideally, lower the potential for further government scrutiny. Since 1988, prosecutors had kept Microsoft staked out as if they were the Gambino family, a trend that would only intensify as time went on.
ACT's initial effort was to focus on interactive television and other broadband network applications as well as low-bandwidth on-line services, a.k.a. Internet technologies. Myhrvold had several pet projects in place, the most realistic of which was the low-bandwidth Internet strategy, outlined in the memo.
Just a month earlier, Myhrvold had been standing with a marker at the white board, outlining different strategies.
"Low bandwidth is the way to make money now," Eller said.
Myhrvold scratched his beard and nodded in agreement. At least Eller thought Myhrvold's nod meant agreement.
The low-bandwidth idea was to connect computers together over the humble telephone line using standard protocols to exchange information in the on-line world. It was, in short, the World Wide Web. Of course, in 1992, the Web didn't exist at Microsoft or anywhere else.
Even so, the opportunity ripe for Microsoft at this moment was to create a standard protocol for the on-line world that everyone could adopt. Microsoft could then own the standards for all subsequent Internet applications--the Next Big Thing.
Gates should have been a natural at this. After all, he was the Standardization King. Before Gates arrived on the computer scene in the mid-1970s, a wide variety of different machines cohabitated, but none of them could communicate with one another, and they had little in common. Gates's first bolt of genius was to unify them with Microsoft's BASIC programming language.
Then, less than ten years later, Microsoft did it again. It brought graphics to the PC and set the new standard with Windows. So it was not a far-fetched idea that Microsoft could lay the same claim now to the Net. The Soft already had the knowledge to bring graphics on-line--Myhrvold said as much in his memo:
I believe that we will see a replay of this situation. There will be
But Myhrvold did not spend the whole memo talking about low bandwidth or Internet standards. He and Gates were also excited about an evolutionary dead end called high bandwidth.
In 1992, Myhrvold was not the only futurist in the computer industry talking about the "grand convergence" of television and computers, a high-bandwidth strategy. Speculation ran rampant about a world with five hundred TV channels, video on demand, online shopping, and interactive games. Massive servers would house all of this information and shoot it down through fast wires into consumers' homes. It all sounded very intriguing. But also very far off in time. This was the Jetson-like world that captivated Bill Gates's geek imagination.
The emerging low-bandwidth Internet could have been a starting place for Microsoft to facilitate that kind of convergence, but low bandwidth simply didn't turn Gates on. It was too ... mundane.
David F. Marquardt, a general partner of Technology Venture Investors in Menlo Park, California, and a Microsoft board member, recalled his amazement that Microsoft was putting so little into the Net. "They weren't in Silicon Valley. When you're here you feel it all around you," he told Business Week in July 1996. He brought up the topic of the Internet at a board meeting in April 1994. Marquardt described Gates's response this way: "His view was the Internet was free. There's no money to be made there. Why is that an interesting business?"
Gates instead would rhapsodize about a world with infinite channel choices, digital art on the walls, and computerized sound, light, and temperature controls throughout the home.
Ironically, these hubristic notions had already placed what could have been the key to success in Microsoft's hands. In early 1992, they had established a project called Homer to look at setting communications standards for the home. Gates would later use some of these ideas for his $53.4 million mansion on Lake Washington. And it was this initiative that eventually gave rise to the Remote Information Protocol.
The idea behind RIP, a concept that Netscape would later realize so beautifully, was to create a graphical user interface that would allow computer users to connect to information providers, much like today's World Wide Web. Microsoft already had an installed base of tens of millions of DOS and Windows users ready to go online. If Microsoft had simply bundled the RIP technology in the next version of its operating systems, the World Wide Web may have evolved in a very different way, leaving nothing for Netscape to create.
Myhrvold gave Eller funding and staffing for the Remote Information Protocol group, and Eller took a handful of people with him from the Homer project. In searching the company to fill the other positions, he wanted to make sure he avoided QVers--Microsoft argot for people who were "Quietly Vesting." Usually they didn't have more than six months before their stock options fully vested. Typically they were already worth a few million, so during their last six months, QVers weren't exactly motivated.
Microsoft was much more lenient about funding and forming groups in the early 1980s. In those days a developer would float an idea by Gates, and if it sounded reasonable, the money followed. Development leads were free to staff their groups with as many people as needed. The problem was, it was still hard to find enough truly first-rate minds to staff all of the groups.
The hiring policy for the company had always been an open call for bright developers. Microsoft had plenty of work available, so when a talented person was spotted, they were hired--they weren't necessarily hired for a specific position.
As standard Microsoft protocol, a dozen lead developers would interview each candidate, and of course everyone wanted the good ones. The head of the Excel group would decide they needed a certain developer. Someone from the Word group would claim they needed the person more. It became a shouting match.
Eventually, Eller emerged with about a dozen people for his group, most of them from within Microsoft. Some had just finished other projects in networking or applications, others were just tired of the groups they were in, and some had been in the old Homer group.
At this moment, Netscape didn't even exist. Sun Microsystems, a behemoth in the workstation market, was riding on hardware revenues from its UNIX business. They weren't focused on software--certainly they had not yet commercialized Java, Sun's now-dominant interpreted scripting language for the Internet.
So the Net just loomed there like a great white whale on the horizon, and Eller was finally in a position to build what he had always wanted--an interpreted object-oriented scripting language. It had to be lightweight, portable, and secure, just like Sun's Java is today. But the interpreted scripting language was only one part of the Remote Information Protocol. RIP encompassed a number of Internet technologies including a browser, like Netscape's Navigator, compression and decompression technologies, and encryption.
By December, Eller's group was rolling along. One day, Myhrvold caught Eller in the hall to discuss their progress.
"Hey," Myhrvold said. "I had this cool idea for doing image compression. Check this out."
Eller followed Myhrvold into his office. He sat on the leather couch while Myhrvold walked around the room waving his hands in the air, his eyes lit up, his voice giddy.
"You know that shaded computer graphics stuff we've been talking about. It occurs to me you could represent the shading by the electric field given off by an electron charge. Think about it--if you've got a collection of static electric charges, they create this electric field. It's more intense around the charges, and it falls off as you move away from the charges as an inverse square." "Uhh, yeah," Eller said, pursing his lips.
"Well, but you see ..." Myhrvold paused as he went to the white board and started drawing E and M equations. "There's all this machinery for solving electrostatic charge distribution problems, like Maxwell's equations. So you ought to be able to take a given electric field and figure out what the charge distribution would be to yield that particular field. Then you could compress the picture by solving for the electron charge distribution, and then just shipping down the locations of the electrons and what their charges are. Then the machine on the other end can reconstruct the picture by solving the field equation for the charges."
Eller scratched his head beneath the beret. It had been eighteen years since his college mathematics major. He didn't remember any of the equations for doing charge distribution on electrons, and he didn't want to look them up. It wouldn't have done any good anyway. Myhrvold was just throwing out some random and utterly convoluted way of doing the same thing Eller was already going after on a much more direct path.
Talking to Myhrvold was a little like smoking dope. It could give you "insights," but in the light of day those insights often didn't make any sense. Eller walked out of Myhrvold's office reeling and dizzy and looking for food.
How could Gates put this guy in such a powerful position? It's not that Myhrvold wasn't smart--he was exceedingly smart. But Myhrvold was a cosmologist. Cosmologists studied physics, but then they wandered off into the big bang and exploding stars and what happened in the first nanosecond. Among experimental physicists, cosmologists were known to be flakes. While physicists were rounding out to the thirteenth decimal, cosmologists were talking about wormholes and super strings and what happened before the universe began.
Plenty of developers resented Myhrvold. They were the ones who actually wrote code and knew the nitty-gritty. Myhrvold, being a big-picture guy, tended to forget how long it actually took to build something. Myhrvold argued that you could design a new graphics architecture in only two weeks. People like Eller, who had personally spent three years developing graphics for Windows, knew better.
Eller was still trying to regain his balance as he walked down the stairs to check on one of his people.
Everyone in Eller's group was in the same building, though spread apart. Microsoft's rapid growth had filled in all the nearby offices before his group had been fully staffed. Still, Eller worked hard to insulate his developers from upper-management politics, to allow them to focus on what they did best: code.
Ideally, Eller wanted to enable users to have a highly interactive--and very cool--user interface. He didn't want a huge server doing all of the processing work. If the server was doing it all, the server quickly became bogged down, and users lost their interactivity. The idea was to be able to take an application--a shopping form, for instance--download most of it to the user, or client, so that it was running on the PC, then let the user play with the form and customize it.
Eller had assigned a handful of people in his group to explore which interpreted scripting language they should use. Two people were looking at the C language, and one person was looking at transport layers, or how two computers would connect and communicate.
Two of Eller's people were working on hacking up Visual BASIC, and two others were working on the Forth language. The Forth language was originally invented by physicist Charles Moore in the late 1960s. It was a widely used, but little known, computer language, and it was very easy to port to different machines. It was tiny and fit in about 3 kilobytes of memory, versus today's systems, which require megabytes, a difference of five orders of magnitude.
Eller concluded that it was crucial to write an entirely new interpreted scripting language, and the issue was still a bit of a sore point with both Myhrvold and Gates.
"The last thing this company needs is another fucking language," Myhrvold had said.
Eller saw Myhrvold's mouth move, but when the cosmologist talked, it was Gates's voice that Eller heard. Gates wanted all of the languages to be the same--BASIC. After all, Gates had written it. But despite Chairman Bill's monument to code, the computer world needed to move on.
And Eller knew BASIC wouldn't work in this application anyway because the scripting language in RIP had to be compact and object oriented. So Eller's group forged ahead.
By the spring of 1993, the RIP group had made significant progress developing their lightweight interpreted scripting language. In fact, if they had stayed on track, they could have gotten RIP bundled into the then upcoming version of Windows 3.11, which was released in November 1993. What would Judge Penfield Jackson have made of that?
But things were already starting to disintegrate.
Myhrvold, who had earlier agreed with Eller that low bandwidth was the way to go, now seemed bored with the project. Instead, he was putting all his energy into the high-bandwidth interactive TV project.
Matters weren't helped any when, four months into the project, a buddy of Myhrvold's named Craig Mundie was brought in to be Eller's manager. Mundie, in his forties, had been the CEO of Alliant Computer--before the company went bankrupt. Alliant was a supercomputer shop that had focused on high bandwidth.
When Mundie came on board he promptly went to Eller's office to check on the RIP team's focus. This was December 1992, and he was not pleased to find Eller and his troops still working on low-bandwidth solutions.
"You're going to have so much bandwidth in the short term, it won't even be funny," Mundie told Eller. "Furthermore, wasting time on 9600 baud lines is just stupid. And last, but not least, you need to get with the program. You better get the message to your troops, 9600 baud is bullshit. You should be focusing on broadband!"
"Yeah, sure," Eller lied. "We can have a joint meeting, and I'll review the troops. No problem, Craig."
When Mundie came back a week later, he was surprised to find that Eller's troops hadn't got the message. They looked at Mundie with glazed eyes as he talked about how broadband was going to be all over the planet. They stared blankly as Mundie explained how people wouldn't need low bandwidth and that everything Eller's group was doing, squeezing bytes here and there to fit on a 9600 baud line, was just a waste of time.
Eller said nothing, and the entire group gazed back at Mundie with a placid look that, deep down, silently expressed, "Screw you."
Mundie followed Eller back to his office for a closed-door post-meeting and demanded to know why Eller's troops weren't on board.
"Well, you know, Craig, they're a hard bunch of guys, and they're kind of focused on the low-bandwidth strategy."
Eller paused for a moment, adjusting his grip for the proper spin. "And I don't know that we need to diffuse that thrust. I mean I'm sure that broadband is going to be great stuff in the future ... and I'd be all over it if I weren't all over this project. But I am."
That's when it became apparent to Mundie that Eller was unmanageable.
Eller had been at Microsoft for more than a decade. He had spent three years in the "death march" to develop the Windows graphics subsystem, GDI, the graphics still used in Windows 95 today, He had also created Pen Windows, Gates's pet project at the time.
But as Microsoft grew, even top developers like Eller no longer had a direct line to Gates. In the early days of Windows 1.0, Eller had squabbled and sparred with Gates regularly. People on the Windows team had gone to movies together and worked till 2:00 A.M., happily hacking as the rock music blasted. Now there were vice-presidencies growing like kudzu. There were too many people to placate, to convince, to cajole.
Plowing through Microsoft's weed-infested org chart, Eller knew the only way to keep RIP alive was to take it straight to Bill.
A recruit party was coming up at the Seattle Museum of History and Industry, and Eller was responsible for chaperoning his newbies to the function. Recruit parties, according to a recent Gates memo, were no longer optional. This was a command performance.
Once upon a time, a recruit party consisted of a case of beer and pizzas in cardboard boxes. The developers would all sit in the lobby of the Northrup building talking to Gates about how great the world would be with their new software. But even now, with oldtimers chumming up acne-faced kids over prawns and chardonnay, Eller still knew that, once the caterers started cleaning up, it would be an opportune time to talk to the boss.
After Eller's new hires had gone, he walked up to Gates in the "Keys to History" exhibit. The caterers were breaking down tables and chairs, and Gates was standing at the buffet chewing on a shrimp the size of Manhattan.
"Hey, Bill," Eller said.
Gates acknowledged Eller with a nod.
"I think we've got a good group here," Eller went on. "Especially that new kid who's gonna work with me on RIP."
Gates nodded again, but the richest man in America seemed otherwise engaged.
"You know," Eller said. "I don't think you're giving enough attention to low bandwidth."
Gates continued chewing, and Eller picked up a shrimp himself. Gates looked at him and pushed the bridge of his glasses back up onto his nose.
"We have low bandwidth today," Eller continued. "Everybody has a modem. People can exchange information at 9600 baud. We don't need to wait for fiber. This way it's an evolution where everyone keeps their existing software and computers. We should do that now."
A long silence hung suspended between the two. Eller knew Gates had heard him. But Gates gazed off in the distance, seemingly oblivious to the Willits canoe, to the black and white stills of early Seattle settlers--and to Eller's point.
"Uh huh," Gates muttered.
And at that moment, Microsoft missed the technology boat. Millions of future Web surfers bearing the Microsoft logo simply turned and paddled back out to sea.
In the early days of pizza and beer, Gates would have been animatedly talking ideas with a developer. He would have been all over the issues, saying, "Why will this be more important than that? Is this really the right thing?"
Now Microsoft had become so big that Gates could no longer focus. Eller realized that Gates wasn't going to encourage anyone to support RIP. But Eller was tired of fighting, tired of persuading, tired of convincing. He walked away from the party disillusioned, but also realizing that maybe this moment of blindness was not really so unusual. Knee-jerk reactions and panic had always been a way of life at Microsoft. Working for Chairman Bill had been like white-water rafting, not ocean cruising. There were so many other near misses and episodes of dumb luck that the public and Microsoft investors never knew about. Now that the company had turned into its nemesis and become just another lumbering giant like IBM, Eller wondered how long market momentum would continue to carry it. Then again, Microsoft had three very substantial attributes to see it through: a lush array of laurels to rest on, enormous cash reserves, and a bone-crushing hold on the ultimate core asset--total control of the operating-system business.
If the past was any indication, Gates would be able to leverage that asset, wielding it like a club if need be, well into the future. And woe betide any competitor who got in his way.
© Copyright 1998 Marlin Eller and Jennifer Edstrom
Back to the top