__ Microsoft's head honcho for Windows 2000 seeks perfection. It's a lonely crusade. __
Before the lights go out, before the money suddenly disappears from your bank accounts, before planes fall from the sky, before Y2K rears its ugly head, Microsoft will bring Windows 2000 to market. Honest. After years of missed launch dates and broken promises, recrafted strategies and skirted deadlines, the next great leap in Microsoft operating-system technology is at hand. As of September, company spokespeople and executives were saying (strictly on the QT, of course) that W2K would ship before December 31. Now for the real news: The release will come not so much with fist-pumping enthusiasm as with a tremendous sigh of relief from Jim Allchin, the Microsoft senior VP who has been the biggest champion of the Windows franchise in recent years.
The market first expected this major upgrade to the Windows NT OS back in 1995. Five years later, you'd think Microsoft would have ironed out the product. And yet, by the start of the fourth quarter, the launch date remained uncertain because W2K still had bugs. Which is actually no great surprise. Any complex software the size of Windows 2000 - there are 29 million lines of code in the client version alone - will have bugs. But Allchin says many of the bugs, and the resulting delays, were caused by a "sloppiness" that has forced a reassessment of the development process.
"When you're moving so fast, it's easy to let leakages happen," says the 48-year-old commander of the consumer and enterprise versions of Windows as well as the streaming-media group. "You're driving the team: 'Make progress in this particular area.' And they're saying, 'I don't quite have the interface I need on this other piece.' We're driving 'em, so they change it." But overlapping changes cause cascading errors, better known as bugs. "And it can be quite depressing," says Allchin, "with so many bugs that everybody knows exist."
Windows 2000 (renamed last fall from NT 5.0) was originally intended to be much more than an upgrade: It was supposed to unify the corporate OS - Windows NT - with the consumer OS - Windows 9x. At last, the DOS technology underlying the consumer operating system would give way to the more stable NT kernel. While the unification would go unnoticed by most users, it was meant to be a show of faith to developers. No longer would they need to write separate code for consumer and corporate software.
Alas, W2K won't meet that promise. Earlier this year, Microsoft admitted that the common code base is still years off. Even the DOS-based next-generation consumer OS - codenamed Millennium - has been pushed back to 2001. Don't expect Neptune, a consumer OS based on NT, until at least a year later. (Millennium and Neptune are on parallel tracks within Allchin's groups.)
By now, of course, missed ship dates and buggy releases have become the MO at Microsoft. Analysts, beta users, and the media seem to have built delays and bugs into their expectations. Reviews suggest that W2K combines the stability of NT with the best of Windows 98 - power management, for instance. As a result, the product should do well at market. International Data Corporation projects Microsoft's share of server units sold worldwide to increase from 38 percent in 1998 to as much as 44 percent in 2003 - thanks in large part to W2K.
Beyond the numbers, though, is a metastory. It ends with the realization that, despite the sell-through projections, Allchin - a nine-year veteran of Microsoft and an indefatigable perfectionist - fell short of his goal. But the most interesting part of the story, as usual, is what happened along the way.
Allchin never sought the spotlight. He was nonetheless thrust into it last winter as part of Microsoft's botched attempt, during its antitrust trial, to discredit a government witness with an amateurishly doctored video. The incident made international news and undermined the software giant's already battered image. In the process, Allchin - a man praised by colleagues and peers for his intense pride and integrity - was sideswiped.
At the trial, government witness Edward Felten, a Princeton University computer science professor, claimed to have written a program that could, contrary to Microsoft's claims, uninstall the Internet Explorer browser without adversely affecting the OS. To refute Felten's point, Microsoft attorneys and marketers pushed forth Allchin, accompanied by a video demonstrating that Felten's uninstall did indeed hurt the performance of a computer running Windows.
While Microsoft's claim may have technically been true, the video was a fake. As government attorneys quickly pointed out, the tape had been cobbled together from footage of several different machines. Gates called the incident an embarrassment. No one was more humiliated than Allchin.
Months later, when I asked him in an email to explain what happened, Allchin dismissed the ordeal - in nine passionate, lengthy paragraphs. "I had people come to my office and basically break down, saying it killed them to know my integrity was questioned because of [their] mistake," he wrote. "Personally, I let this go a long time ago. People were trying to save me time. In retrospect, I should have done it myself. The sad thing is that this is such a small point: an interesting newspaper/dinner topic, but not the point at all.
"The question," he continued, "is whether we did something good for consumers and whether our innovation of integration offered compelling value. Innovation to win customers is what business is all about. We innovated through unifying the browsing experience for local files, server files, and Internet files. I stand by this."
Aside from hurting his reputation, the trial distracted Allchin from his long struggle to shepherd W2K to market and, in a larger sense, protect the Windows flagship inside Microsoft. To those ends, Bill Gates and Microsoft president Steve Ballmer earlier this year handed Allchin command of both the consumer and business divisions of Microsoft. Now, with W2K poised to launch, Allchin confronts a world in which Windows is waning as a power - even, if you consider a September meeting held in San Francisco to tout the company's cross-platform ambitions, within Microsoft itself.
Given Allchin's personal history, this is an almost poignant twist of fate. The world of the Internet, where millions of people share resources across a decentralized web of computers, is a world Allchin envisioned as a doctoral student at the Georgia Institute of Technology in the early 1980s. At a time when freestanding minicomputers like Digital's VAX machines dominated college campuses, Georgia Tech faculty, with Allchin's help, were developing an experimental computing environment called Clouds. Allchin's PhD thesis laid out an architecture for Clouds that would let computers share resources. Indeed, it was a preview of distributed computing - where the Net itself becomes one big computer - as envisioned today in such software products as Sun's Jini.
Allchin went on from Georgia Tech to help design the Vines networking software at Banyan Systems, which, at the time, was the PC networking leader. When Microsoft executives first approached Allchin about a job in 1989, he wasn't interested - frankly, he thought Microsoft's products sucked. The company's reputation, he says in the soft twang of a transplanted Southerner, "was truly - and this is just my opinion - truly quite bad. Microsoft was not known as a company that built quality software.
"I kidded them, saying, 'You brought us DOS and then you followed it up with OS/2 - both of which I don't have much respect for,' " Allchin remembers. "At Banyan, we tried to use OS/2. It was just so bad! It wouldn't print. It was just not architected right at all."
Obsessed with Novell's competitive threat, Microsoft considered Allchin a key to the future of its networking-software strategy. The company courted him intensely for nearly a year, in an effort that included the ministrations of both Ballmer and Gates. We're really serious about mending our ways, the Microsofties insisted. In the end, they bagged their man with a dare. What bigger challenge could there be for a programmer than making Microsoft software work? "Bill said, 'No matter how good your software is someplace else, you can probably touch more people by coming here. We'll give you the resources, and we'll build it here,'" Allchin recalls. "Maybe I'm sappy, but that got me."
Once inside, Allchin aligned himself with Windows. In his first six months, fresh with his new religion, he convinced management to dismantle the 350-person programming group responsible for developing networking software based on OS/2, the operating system developed jointly by Microsoft and IBM that was supposed to succeed DOS. "I know I probably upset a few people during and after that," he says.
He worked slowly to grow networking muscle inside Windows NT, Microsoft's then-nascent server technology. With the release of NT 4.0 in 1996 - six years after Allchin came on board - Microsoft introduced a solid competitor to Novell's NetWare.
In the beginning, Allchin held on to his dream of distributed computing, which eventually evolved into the Cairo project - a pre-Web attempt to build a brand-new OS that would include new ways of browsing for information across platforms. "I took a team, and we tried to lay out a future vision for what could be done with deeper distributed computing," he says.
Cairo was seen as a failure, because it never shipped - although key pieces of it, including the user interface, would resurface in both Windows 95 and NT. "It was probably overly ambitious," Allchin now says.
By then, Microsoft had risen to dominance due to its singular focus on volume, and Allchin's mind-set had changed. Critics contend that the volume strategy, designed to generate continual upgrades while locking out competitors, lets Microsoft bring to market late, buggy products - with fewer features than promised - and still almost always do well.
Allchin quickly embraced the strategy as his own. "Windows is the volume model, and Jim became a big believer in volume," says Dave Mahoney, the Banyan Systems founder who hired Allchin out of graduate school and subsequently lost him to Redmond.
Come 1997, the Internet had taken its hold on the market, and questions arose about the future of a closed-architecture system. By this point Allchin had become a true convert: He was one of the most ardent believers in and supporters of the Windows franchise in the face of an increasingly open source world. His vocal support of Windows placed him opposite the comparatively Net-centric Brad Silverberg in a heated strategic debate about cross-platform development and the fate of Windows.
Silverberg, another senior VP at Microsoft, had long been involved in management of the consumer Windows group and had orchestrated the launch of Windows 95. In the wake of Silverberg's success, Gates asked him to spearhead the struggling Internet group, which then was engaged in the early stages of its browser war with Netscape.
The group was intent on matching Netscape's development across three operating systems: Windows, Mac, and Unix. Silverberg's team members became outspoken advocates of rival Sun's nascent Java platform creed, "Write once, run anywhere."
Incensed at the growing power of an idea he violently opposed, Allchin sent a now-infamous email to Gates and Paul Maritz, the longtime head of developers at Microsoft. "Windows," Allchin wrote, "is in the process of being exterminated here at Microsoft. I consider this cross-platform issue to be a disease within Microsoft."
It was a critical moment at Redmond - an issue that called for intervention from Gates himself, who decided the matter by siding with Allchin. According to The Wall Street Journal, Gates, at a meeting with members of Silverberg's group, screamed, "Windows is what this company is about!" A few months later, Silverberg departed on a sabbatical from which he has yet to return.
"Although there is always debate within Microsoft about specific implementations of technology," Gates told me in an email, "the Internet protocols and infrastructure were always going to be in Windows NT. It's really about combining the power of both the Internet and Windows."
Last winter, it appeared that Silverberg might return when, according to Silverberg himself, Gates and Ballmer offered him control of the consumer Windows group. But when he said no, they turned to Allchin.
Allchin now plays down his outburst, even while acknowledging his fear that Microsoft was close to losing the biggest weapon in its arsenal. "Email conversations are a little more flamboyant," he says. "You can't raise your voice, you can't gesture your hand, and so you sometimes capitalize words, you sometimes italicize something. And you often will just take an extreme view to try to get people's attention. But was I worried? Yes. Deeply.
"We were taking code from the new versions of Windows and moving it back to installed base versions and not charging for it sufficiently, in my view," he says. "And there were mail messages about the move to clone the Windows APIs in Java, which I strongly disagreed with. I saw no reason to do this. It would do nothing but create an alternative to Windows that would confuse customers."
And so Allchin won the battle - with tactics quite at home inside Microsoft, a company famous for raging internal conflagrations as part of its corporate culture. Supporters of opposing strategies square off for weeks at a time, peppering each other and their bosses with emails and memos. Only the strongest arguments are expected to survive. It's an environment in which Allchin, with his directness, has flourished.
"Jim has been a key contributor to the business since he joined Microsoft," Gates responds when I ask about Allchin's role and character. "His initial charter was to drive the company's networking-product strategy. Since then he has led the development and marketing efforts for Windows, including Windows NT/2000, Windows 98, and streaming media. He's a brilliant technologist, visionary, and a strong leader."
Allchin may have saved Windows, but at what cost? The onetime champion of a distributed, decentralized computing system now oversees, in W2K, a monstrous piece of software that critics like Sun CEO Scott McNealy deride as "The Big Hairball."
Allchin says much of the bulk has been added in recent years to support functions that Microsoft couldn't have anticipated early on, and he balks at the suggestion that W2K seems about as far from his early dreams as he could have traveled. Distributed computing is still on his mind, he says. "There are some debates that we've had for years, and we're still having for years. Cross-platform is one that came up, and it was twisted six ways to Sunday. Cross-platform can mean different things. People can change their views.
"Distributed computing is about having computers talk to each other," Allchin continues. "Each vendor is responsible for supporting standard protocols so that the different boxes can talk to each other. My energy has been focused on intelligent machines communicating with each other. I am a hardcore believer in distributed computing. Competitors might try to position Windows as about desktop computing. That is totally ridiculous today."
While still passionate about big issues, as the year comes to a close Allchin is most concerned with making a go of W2K in the short term. Fortune 500 IT managers don't like the bleeding edge, and bugs, of course, get in the way of multimillion-dollar purchasing decisions. Additionally, the success of W2K's predecessor, NT 4.0, could actually hurt the new product. "On the server side, NT has just been fantastic," says Ziff-Davis chief analyst Aaron Goldberg, noting that because NT has worked so well, many won't upgrade. On the client side, uptake may come faster, but "Linux is sneaking around the corner."
The delays may have hurt Microsoft most, in fact, by allowing Linux to make much heralded inroads with PC makers. Even staunchly Microsoft-loyal Dell Computer announced in September that it would begin offering Linux to customers upon request.
Still, Allchin is not without his allies. All the major manufacturers have committed to carrying W2K to market. Micron Electronics, for example, plans to implement it most significantly in high-end workstations and servers. "I was pleasantly surprised at how strong it was," says Micron VP Michael Gale.
"The more features they bundle, the more of a deal the product seems to be," says IDC analyst Dan Kusnetzky. "And that locks out the competition by forcing them to come to market with the same features at the same price."
But Kusnetzky also sees a potential hurdle at the high end. "Microsoft's own business model is getting in the way. It sells a commodity product, and every 15 to 18 months builds an upgrade," he says. "Every two releases or so, the requirements of the software have gotten so big the machines have to be replaced. If I've bought a three-quarters-of-a-million-dollar multiprocessor server, I'm not going to replace it that often."
To meet the needs of a changing environment, Steve Ballmer and Paul Maritz announced in mid-September a strategy acknowledging that not every computer connected to the Internet need run Microsoft software. It's a position notably similar to the one held by the firebrands in Silverberg's Internet group before they were slapped down by Gates. Appearing before an audience of press and analysts in San Francisco's Sheraton Palace hotel, Maritz announced, "People don't have to buy into our whole programming infrastructure to interoperate."
Allchin, as it happened, was not present. Perhaps because the message didn't align with the strategy he'd been hawking. Or maybe because he was in Redmond, killing bugs. Either way, he was sounding a bit defeated several weeks before launch. "It's hard, because I'm super proud of what we've accomplished," he said. "But at the same time, I can be depressed at setting high goals and not quite reaching them. I can tell you so many things that we should have done better. And if you talk to my team, they can, too."
So while Microsoft contemplates the post-Windows future - a future once imagined by Allchin and then abandoned for the good of the company - one has to wonder what lies ahead for the man behind W2K. Some say it's impossible to accomplish anything inside Microsoft these days: too many internal battles, too little sense of urgency, and an institutional fatigue from the antitrust trial. Any vision with legs is quashed.
"It's a huge Goliath, and changing a culture is very difficult based on dictate," says an executive with a major PC manufacturer. "I don't necessarily see the incredible market-share-grabbing energy that in my opinion ended with NT. There's a stream of management that's not focused enough. They need some fresh blood. You can't ask those inside to reengineer from inside, because they're too wedded to what they've learned."
True to how former colleagues describe him - as extremely honest - Allchin says upper management is working on any number of such problems. "We are reengineering ourselves," he says. "It's just a long grind."
Is Allchin experiencing a crisis of faith? Or has he become just another complacent Microsoft millionaire? One email exchange away from the fate of Brad Silverberg - on permanent sabbatical? In late August, rumors began to circulate that Allchin was on his way out. And while he may pay such gossip little mind, there are indications that he could certainly find enjoyment in life beyond Microsoft. For starters, he and his wife, Catherine (a former employee of Waggener Edstrom, Microsoft's PR firm), had their first child this summer. On top of that, Allchin recently finished building a home studio for his other passion, the guitar.
But in the midst of sweating over the little things in the days before launch - like the time it takes for his computer to wake up after he sets it to hibernate - he remained all business.
The last time I saw him, Allchin sat at a conference table with 22 programmers, trying to motivate them during a weekly bug report on the workstation version of W2K. "Have you guys seen Millennium? Their hibernate is much, much faster," he said. "It's pathetic! You guys gotta see Millennium!"
It was a cheap ploy - playing one development group against another - but the staff took it in stride. After some back-and-forth as to whether there was something unique about Allchin's Toshiba notebook, they promised to have a look. "Well, if you get it in the 25-second range, we won't be totally humiliated," he said with a small laugh.
For Jim Allchin, life at Microsoft has been a series of compromises - some smaller than others.
