Resolve and Fortitude : Microsoft's ''SECRET POWER BROKER'' breaks his silence (27 page)

Years too late and fortunately for us, the program failed miserably. Most analysts believed our Windows platform was too mature and entrenched, presenting Apple only with a slim chance to gain at our expense. The real reason the late attack failed? First, Macs were still overpriced compared to IBM PC clones regardless of who built and sold them. Second, Apple picked small, underfunded companies to build Mac clones. Her partners were further disadvantaged when Apple kept essential information close to her chest and provided only scant lip service supporting them. The Mac had a formidable application software base by then. In enlarging its manufacturer, base share it should have been rapidly won increased market share. Instead, all of Apple’s licensees painted a dreary picture as I visited them, trying to license our Mac apps. As a company steeped in proprietary principles, Apple never valued the profound advantages her licensees could have brought to the table.

After Steve Jobs was reinstated as Apple’s CEO, he simply discontinued the program by not extending the deal to the new iMac architecture he invented. Bill was very skeptical about its spiffy design. More aggressively priced than before ($1,300 per unit), it nevertheless sold eight hundred thousand units in the first twelve months and took market share away from Windows. While this hurt some and started a still ongoing trend, the greater danger for us loomed surprisingly in the UNIX corner.

SOFTWARE MINUTEMEN

“Linux is a cancer.”

Because it attaches itself in an intellectual property sense to everything it touches, as Steve Ballmer once formulated. Until now my OEM business wasn’t much affected by that emerging phenomenon. But in late ’98, its threat to our desktop business took a dangerous turn. There were indications that the free-of-charge Linux OS was being bundled with recycled $500 PCs in developing countries. Then plans emerged to preinstall it on low-powered notebooks
36
and Oracle’s planned N-cube system priced under $200. Scary scenarios for us, with Windows average price being $55 per unit.

Let’s quickly examine why Linux attracted so much speculative attention and support from top-notch companies. As the name and letter combination suggests, it derived from UNIX. A misleading assumption, but in itself the right evocation! Its inventor was Linus Torvalds, a genius of a Finnish programmer who had in ’91, while studying computer sciences at the University of Helsinki, succeeded in creating a UNIX OS clone from scratch. Written mostly in the C programming language, he adapted his clone to the Intel 80386 CPU. However, he was not the first person to achieve such a feat; Dutch professor Andrew Tanenbaum had preceded him, naming his cloned version MINIX. While the name MINIX suggested a small system (mini-UNIX), the name Linux derived from Torvalds’s first name, Linus, and as such is a nice play on both his and the UNIX name.

Originally, both contained only essential bare-bones components to make them simply functional. They were operated with a command-line language, comparable but harder to use than the ones found in MS’s or DRI’s DOS versions. Both releases were distributed in source code, unlike the distribution of Windows using binary code. This offered students and hackers around the world an opportunity to study them in-depth and modify or enhance them con gusto. Torvalds’s implementation had a monolithic architecture allowing multithreading and contained a task-switching feature. The only criticism of his implementation: tying his 10,200 lines of code so specifically to Intel’s 386 did not bode well for portability to other CPUs. MINIX, at roughly the same size, was a more modular system, making a later attempt at portability easier. After exploiting both, the hacker community judged Torvalds’s implementation as more complete, refined, and state-of-the-art. Therefore, Linux eventually found its way into everyday use while MINIX remained a classroom tool.

Torvalds had offered his code royalty-free. Could developers redistributing Linux, after making enhancements, at least make money from the handiwork they contributed? Enter the world of Richard Stallman, another genius of a programmer and probably the world’s most outspoken free-software ideologist. When working at MIT’s
37
artificial intelligence laboratory, he became fed up with the stringent nondisclosure agreements he encountered—similar to what the rest of industry was using—and decided to free the world from such restrictions. In ’83 he published a Karl Marx–style manifesto declaring software to be free merchandise. Free as in freedom. Stirring the American soul as a revolutionary thinker, he subsequently went beyond his original idea of unchaining software from commercial constraints. Advocating any software code should be allowed to be modified by everybody, he effectively outlawed the sole distribution of binary code. Going beyond, he radically demanded all resulting code from derivative work on royalty-free software to be dispersed in the same spirit and terms—without dreaded royalty shackles.

By ’91 Stallman had published a second version of his General Public License (GNU
38
GPL). It put his intent into crisper legal terms. Torvalds, hesitant at first, followed Stallman’s philosophy and agreed to license Linux under the newest GPL. Later, industry observers considered this move essential for the breakthrough his OS kernel eventually achieved. For profit, MS was not amused. Steve labeled Linux a cancer, because all later improvements were in the public domain and could not be explored monetarily.

Linux and its add-ons, freed from royalties and with no other strings attached, woke up dormant talent and created novel opportunities for fame. Think about developers getting hold of it and in their spare time adding significant improvements. Acceptance accolades and fame within their community occurred at Internet speed. Observing the trend and hungry for the same recognition, hundreds of thousands followed trying to outdo and outimpress each other. Humanity’s competitive spirit at its best! Suddenly, working on Linux was fashionable and cool and its minions prolific! Soon the merry army of authors gleefully rebelled against the established fortresses built by the likes of Bill Gates, Larry Ellison, Rob Glaser, and other billionaires of the software industry. No need to throw a pie in their faces anymore just to gain attention; here was an effective, exciting, and peaceful method for actually dethroning them.

In a short amount of time, the legions of volunteer worker bees attained astonishing breakthroughs and software stardom in progressing Linux protuberantly faster than Windows. When the first complete version was released in March of ’94, the system had grown to 176,000 lines of code. A year later, it doubled in size, and by January of ’99, the system contained nearly two million instructions. Today, as I write this, the system comprises of ten times as many. Developing it commercially from scratch is estimated to cost nearly 1 billion US dollars. The volunteer movement initiated by Stallman and fueled by Torvalds’s release had been a truly amazing—hats off—ride for the software minutemen.

Smartly and stringently, Torvalds, now considered both a hero and a bottleneck, managed to maintain control over his various core releases. The gurus surrounding him restlessly wanted to move faster. Torvalds’s demand for quality and his scrutiny for robust design held them back. Keeping careful control of the process, his fastidiousness succeeded, rewarding him not only with community recognition but also with speedy client feedback.

By ’96, Linux had become an established and reliable server product. In ’99, when IBM jumped on the bandwagon, releasing mainframe patches, Linux gained enterprise-class computing honors. Sidelined, our own server group envied Linux’s success! Against initial odds, it had been ported to a variety of other architectures. My own group, with no urgent focus on selling server software, so far had encountered limited Linux competition on desktop PCs or notebooks. There were consistent rumors of Linux finally coming of age and becoming easier to operate. Installation and tuning it nevertheless remained no job for novices. The tools available were impressive yet harder to use than Linux itself. Suddenly, though, a change was on the horizon, as commercial outfits jumped into the fray alongside the volunteer network. Hewlett-Packard, IBM, Intel, and the Linux distributors were eager to expand its use into my group’s core territory, making me mischievously wonder how to compete with royalty-free software.

From its first stirrings, we closely observed the free-software movement. Bill could not stop himself from exchanging barbs with Stallman in public and vice versa. How could a bunch of hackers deliver acceptable, quality code? MS was proud of her keenly managed developers, her smart testers, and her adopted quality standards. Initially, we regarded the volunteer workforce as a fleeting anachronism. Before long, though, these deceptively disorganized gunslingers were tossing out consistent surprises. As human resource managers know, money is not the greatest or most enduring motivator; I actually believe it is none at all. Reaching guru status within the community was a much more powerful incentive—similarly observed inside MS. With Torvalds’s iron enforcer fist in the background, which nobody dared to challenge, this archaic system produced excellent software—though for the moment only for servers.

To make Linux successful on consumer PCs, the ease-of-use aspect needed addressing. Oddly, and luckily for us, the volunteering experts were so far not terribly keen about spending time remedying that obstacle. Loving an intellectual challenge, they had no problem operating Linux effectively. Nobody had ever made it easy for them; why couldn’t you assume a certain skill level at the consumer level? Or, quoting David Eisenberg, “The typical end-user sees the computer as a means to an end… A typical Linux user, on the other hand, sees the computer as an end to itself.” This made ease of use a lesser priority for the volunteering gurus.

The danger of a mind change remained. It would only take one serious software company to produce a consumer-friendly version of Linux. We estimated the effort to be in the neighborhood of thirty to fifty man-years. Cloning Windows completely would take more effort and manpower, but there was no reason to believe that it couldn’t be accomplished in less than two years. If a Linux-based Windows clone managed to hit the market, its impact would be severe and, according to several experts, take significant business away from us. We had to win two battles: Limit the damage Linux was doing to our server business, and make sure the Windows desktop franchise would not suffer irreparable harm. Don’t rest; move Windows forward by adding complexity through innovative features. Making it a progressing and therefore harder-to-attack target was exactly what our carefully considered counterstrategy defined by Bill called for.

In November of ’98, we published a comprehensive Linux analysis for internal MS considerations. It managed, though, to wend its way to the Linux community. Its spies were everywhere! The volunteers eagerly added their criticisms and published the research paper on the Internet, naming it the “Halloween document.” The author classified Torvalds’s release policy as a hindrance to Linux’s success. Linus only controlled a small portion of the code, the so-called kernel. The rest, added later by other entities, varied and therefore led to several versions of Linux being marketed. Companies sinning in that fashion were Linux’s main distributors Caldera, Red Hat, and Apache. As they tried to differentiate their offerings that way, they kept ISVs guessing if their applications would run on all published versions.

Along with their varying core offerings, most Linux distributers bundled application packages like office productivity applications and databases in addition to browsers and a variety of interfaces. All of them at no charge! Major software and hardware companies were now involved, helping Linux to succeed. Helping the enemy as we labeled it! By ’98 the estimated user base of Linux exceeded 10 million systems, with more than 98 percent of them deployed as servers and the rest as developer workstations. The report found feature and execution speed parity between Linux and other UNIX systems—amazing. Successfully chasing taillights in the fog of the server and workstation market, Linux had become the undisputed UNIX-class volume leader.

Could this success be repeated for the consumer desktop market? There was no doubt whatsoever it could, especially with the complicity of companies like IBM or Hewlett-Packard. The hypothetically devastating impact kept Bill and me up at night. The substantial support the esoteric but expanding community was now enjoying made MS its prime target. A more sovereign and efficient tactic, perhaps, than trotting through the mouse maze of the courts and filing expensive and ill-founded antitrust claims! Inasmuch as the minutemen rose up against the English, we had to expect that the software revolutionaries were out to obliterate Windows, our crown jewel.

On the face of it, Linux was free of charge. A misnomer. Distributing and supporting it, however, required funding. Windows came with free support supplied by us, our OEMs, or both. The nexus of Linux support services was where its distributors derived their revenues from. Like MS and other for profit software companies, Linux distributors were listed on the stock market. It was the same story on a different slice of bread; free lunch had always been more expensive than advertised. For a simple desktop system, its support fee, including the right to use an information database, today ranges from $40–$50 annually. For servers the fee can reach several thousand dollars. Assuming a desktop PC system lasted three years, a user would then typically—over its lifetime—pay close to $150 for support. Assuming OEM prices for Windows remained stable, this compared unfavorably to—support included—$55 for consumer Windows and $100 for enterprise Windows NT. Stallman’s grand vision of free software for non-OS-literate consumers was not gifted to humanity as touted. It’s utility and zero sum pricing remains deceptive!

Maintaining such a system was yet another less-than-frictionless task. Installing device drivers for new devices and making them operable often required expert knowledge. While the Linux community delivered them in their dependent fashion, their quality and reliability were not as solid as Torvalds’s own code. Compared to Windows, Linux was also hard to install. There was no plug-and-play technology in Linux yet, another prerequisite for the successful launch of a consumer-friendly version, but again an easily developable feature.