Programming Windows: Happy Halloween (Premium)

“God, how deep is this crater?” Jim Allchin asked.

He wasn’t being rhetorical: After winning the battle for the soul of Windows in 1998, Allchin had emerged as the vice president of Microsoft’s new Platform Group, and he was the most senior executive at the firm responsible for Windows. Both Windows: Allchin oversaw the DOS-based Windows family as well as Windows NT, and he was spearheading an effort to combine the two into a cohesive, single offering. It would be called Windows NT 5.0.

But NT 5.0 was late. Really late. By mid-1998—the time of its original planned release date—NT 5.0 had grown into a monster of project that consumed over 40 million lines of code as more and more features were added by its eager engineers. Over 6000 Microsoft employees were working directly on NT 5.0. But the delays were triggering related delays all over Microsoft, because so many other product shipments were dependent on NT 5.0. And they were presumably triggering some angst with Allchin, who had previously failed to ship Cairo, a planned major upgrade to the original NT. He had to get this one right.

NT existed because Microsoft CEO Bill Gates had wanted a “Unix killer” a decade earlier, a portable OS platform that could run on non-Intel chipsets in case the industry went in a different direction. This ultimately didn’t happen, on the PC client or on servers, but the world’s leading Unix vendors of the early 1990s—Sun, IBM, and HP—all ran on very high-end proprietary hardware. And as NT finally shipped and evolved over time, it started eating away at Unix’s hold on the server market. First with workgroup computing and then with intranets and Internet servers.

But Windows NT 5.0 was going to finally blow past Unix and networking rivals like Novell by offering rock-solid reliability, resiliency, and failover capabilities. And by providing a full-featured directory service called Active Directory that could unseat Netware. With the release of NT 5.0, Microsoft would firmly put its foot inside the doors of enterprises everywhere. And the plan was to close the door behind it, lock it, and throw away the key.

Consumed as he was by internal battles for the control of Windows and by defeating Unix and Novell, Allchin at first didn’t seem to care much about yet another competitor nipping at Microsoft’s heels. This competitor was both independent and free—literally—and it would eat away at NT from the bottom, just as NT itself was then doing to Unix and Novell. It was called Linux, and it was, for Microsoft, a nightmare scenario.

The Linux story is, by now, well understood: Created as a personal project by Finnish programmer Linus Torvalds, Linux is an open-source variant of Unix designed to run on the same commodity Intel-based PC hardware as both Windows and NT. Before Linux, Unix was expensive and the province of white lab coat-wearing experts. But with Linux, Unix became free and available to all. And throughout the late 1990s, it started experiencing a period of explosive growth as it matured. It quickly evolved into a platform that could threaten Windows on the PC client and NT on servers.

Allchin was at least aware of Linux. And while he downplayed the system as a “handyman’s special” publicly, he ensured he didn’t make the same mistake that Gates had made earlier with the Internet. In mid-1998, he asked an engineer named Vinod Valloppillil to write an analysis of both open source generally and Linux specifically. He wanted to know whether Microsoft should take them seriously.

It should. And the resulting memo read like a Tom Clancy novel.

“[Open source software] poses a direct, short-term revenue and platform threat to Microsoft, particularly in [the] server space,” Valloppillil reported back. “Commercial quality can be achieved/exceeded by open source software projects. The ability of the open-source software process to collect and harness the collective IQ of thousands of individuals across the Internet is simply amazing. Since money is often not the (primary) motivation behind open source software, understanding the nature of the threat posed requires a deep understanding of the process and motivation of open source development teams. In other words, to understand how to compete against open source software, we must target a process rather than a company.”

Open source and Linux, in other words, was more akin to the War on Terror than it was on traditional wars. Microsoft would need a new game plan.

Valloppillil’s memos were much like other such memos of the day: A logical, analytical engineer writing honestly and bluntly about a perceived threat. But two of his memos were leaked in October 1998, and immediately seized on as a major marketing opportunity by Eric Raymond, a key figure in the open-source community. He called them the Halloween Documents and he published them along with his own commentary on Opensource.org. And over time, he collected and annotated several internal Microsoft memos dealing with Linux and open source. And he gleefully racked up numerous mentions in the press. Simply by treating Linux as a potential threat, Microsoft had made Linux an actual threat.

More to the point, Microsoft’s take on Linux was mostly correct: At that time, Linux was “an alternative to/competitor for other versions of UNIX, especially RISC UNIX,” and not so much a competitor to NT. In many ways, it was Linux that would be the Unix killer, and not NT, Microsoft realized.

As important, the firm quickly understood that it had a key asset it could use against Linux and open source: Intellectual property. “No intellectual property protection means that the deep investments needed by the industry in infrastructure will gravitate to other business models,” one August 1998 memo noted of Linux. “Unless Linux violates IP rights, it will fail to deliver innovation over the long run.”

Microsoft would eventually use its IP portfolio to great effect against Linux. But in late 1998, the timing was poor. It was embroiled in a major antitrust case with the U.S. Justice Department, which sought to prevent the software giant from abusing its monopoly position in the PC market. And its grand unification project, NT 5.0, was well past its projected release date. “Every day [we’re] not shipping is a day people can buy something else,” Microsoft president Steve Ballmer told author David Bank.

So Microsoft punted.

Putting an end to the delays and a temporary end to the code bloat, Ballmer instructed Allchin to reign in NT 5.0 in late 1998. Plans for unifying Windows 9x and NT were pushed back to a future release, trimming over 10 million lines of code and significantly simplifying the product. Windows NT 5.0 would now address only the business market, and Allchin instructed the Windows 9x team to start preparing at least one more release of that family of products, a sequel to Windows 98.

As for Linux, the open-source system never emerged as a credible threat on the PC desktop, though Microsoft commissioned an Israeli software firm called Mainsoft to port over Microsoft Office in 1999 just in case. And Microsoft had a small scare years later when PC makers briefly bundled Linux on very inexpensive notebook computers called netbooks. But Linux ultimately did kill Unix on the server, and the firms that once embraced expensive, proprietary Unixes quick embraced Linux instead. In doing so, Linux and the open-source community denied Microsoft a victory of which it had been previously assured.

Microsoft wasn’t done with Linux yet.