Summary
In the early 1990s, the personal computer industry was a Wild West, and operating systems were the territory everyone wanted to conquer. Microsoft was rapidly building its empire, but it wasn’t yet the titan we know today. Microsoft’s dominance rested on MS-DOS, a text-based foundation that was starting to show its age. As Microsoft prepared to usher in the future with a revolutionary graphical interface called Windows 3.1, a threat emerged from a competitor. What followed was one of the most ruthless episodes in tech history. Secret codes, engineered panic, a massive corporate lawsuit, and a smoking gun that would eventually cost Microsoft nearly a third of a billion dollars. The clone that outclassed the original DR DOS 6.0 beat MS-DOS at its own game The conflict began with a company called Digital Research. This company was founded by Gary Kildall, one of the founding fathers of personal computing. Two decades earlier, Kildall had written CP/M, the first operating system to turn a microprocessor into a usable computer, and it became the de facto standard for early PCs. He also pioneered the concept of the BIOS. The company Kildall founded launched a competing operating system named DR DOS 6.0 in 1991. For Microsoft, this was an existential nightmare. DR DOS was completely clone-compatible with MS-DOS, meaning you could swap it into your PCs and run all your standard software without missing a beat. Worse yet, tech reviewers overwhelmingly agreed that DR DOS 6.0 was functionally superior to MS-DOS 5.0. It featured better memory management, file compression, and task switching long before Microsoft could implement them. It was cheaper, faster, and steadily gaining ground with corporate buyers. In fact, here’s an email Phill Barret, a Microsoft executive, got when he asked one of his engineers to review DR DOS: “The most important reason to use ANY version of DOS is to run DOS apps. DR DOS 5.0 runs every DOS app I know. DR DOS 5.0 works successfully with Windows (2.11, Win 386 2.11 and Windows 3.0 and 3.0a). DR DOS is vastly superior to MS dos 5.0. Both have nearly identical features …” Microsoft was in a tight spot. Windows 3.1 was not yet a standalone operating system. It was merely a graphical shell that ran on top of DOS. If a consumer chose to run the upcoming Windows on top of the superior DR DOS instead of MS-DOS, Microsoft’s core software monopoly would crumble. Recognizing the danger, networking giant Novell acquired Digital Research in 1991 and poured massive corporate funding into the fight. Microsoft realized that merit alone wouldn’t be enough to beat this rival; they needed to weaponize the code itself. They did. The poison baked into the beta A fake error with one job — scare you off DR DOS In late 1991, Microsoft distributed a pre-release beta version of Windows 3.1 to thousands of software developers and testers. The goal of a beta is traditionally to locate bugs, but Microsoft hid a completely different kind of mechanism inside this particular release. When a tester tried to install the Windows 3.1 beta on a machine running DR DOS, the installation process would abruptly halt. The monitor would flash a cryptic, alarming error message warning the user that a “Non-Fatal error” had been detected, citing an “error #2726” and instructing them to contact beta support. Other variations of the installer threw up warnings about invalid device drivers or implied that the operating system beneath it was, fundamentally, unstable. To the average tester, the message was clear: Windows 3.1 was too advanced for DR DOS, and sticking with the competitor would break their computer. It was a classic deployment of “Fear, Uncertainty, and Doubt” — the psychological marketing strategy known as FUD. The trick worked perfectly until April 1992, when a software analyst named Geoff Chappell — whose forensic dissections of Windows internals are still cited by developers today — decided to look under the hood. Alongside Andrew Schulman, the programmer and author behind the landmark reference Undocumented DOS, Chappell reverse-engineered the Windows setup binaries. In the machine-level hex dumps of files like WIN.COM and the Windows setup program, they discovered a sequence of heavily scrambled, self-modifying instructions. The program wasn’t checking for system stability at all. Instead, it was secretly interrogating “Interrupt 21h” (the primary API channel through which software talks to DOS) and checking for highly obscure, undocumented internal structures unique to DR DOS. If the code detected the competitor’s unique digital signature, it intentionally triggered the fake error screen. The researchers dubbed this hidden mechanism the “AARD code” because the Microsoft engineer who wrote it, Aaron R. Reynolds, had embedded his own initials into the encrypted signature. To keep this routine hidden from prying eyes and regulatory agencies, the programmer utilized XOR encryption. XOR is a cryptographic technique that scrambled the true purpose of the code during execution. This was self-modifying code, a highly unusual technique for commercial software installation. It’s more typically favored by malware authors trying to evade antivirus software. The AARD code was only one landmine Meet “Bambi,” the disk cache that hunted DR DOS When Caldera’s lawyers finally pried open Microsoft’s internal files, the AARD code turned out to be a single entry on a much longer list. Caldera’s complaint catalogued seven separate technological incompatibilities allegedly engineered to trip up DR DOS — among them the AARD code, an “XMS version check,” a “nested task flag” trick that hooked into an obscure x86 processor flag, and a mechanism Microsoft’s own developers had affectionately code-named Bambi. Bambi wasn’t a cartoon deer. It was Microsoft’s internal code name for SMARTDrive (SMARTDRV ), which was the disk-cache utility bundled with Windows 3.1. According to Caldera’s filings, at least one Windows 3.1 beta buried detection code inside Bambi: if it sniffed out DR DOS, it threw an error and refused to let Windows run at all. Microsoft had locked Digital Research out of beta-testing Windows 3.1 before its public release, so DR DOS engineers had no chance to catch these incompatibilities and patch them in time. Yet internal analysis later showed DR DOS could be made to run Windows perfectly with only minor modifications — proof, Caldera argued, that the breakage was manufactured. The defense that fell apart in discovery Microsoft’s own emails became the smoking gun When the discovery was made public, Microsoft’s public relations team went into damage control. They argued that Windows 3.1 was a complex environment that had only been certified to run safely on genuine MS-DOS. The AARD code, they claimed, was just a safety feature to alert testers to potential compatibility risks. Though Microsoft disabled the error-triggering routine in the final retail version of Windows 3.1 shipped to stores, they left the dormant code entirely intact inside the software. Digital Research quickly issued a software patch to bypass the check, but the commercial damage was already irreversible. | | | A sealed Microsoft Windows 3.1 User’s Guide, which came packaged with the operating system’s set of 6 floppy disks. | The FUD campaign had done its job. Corporate buyers and original equipment manufacturers became terrified that future retail versions of Windows would completely block non-Microsoft operating systems. Coupled with Microsoft’s aggressive licensing contracts, which forced PC manufacturers to pay a royalty fee for every single machine they shipped regardless of whether it had MS-DOS or DR DOS installed, the market space for competition evaporated. Starved of oxygen, Novell abandoned development of the operating system by 1994. The true, predatory nature of the AARD code remained a matter of fierce industry debate until a small company named Caldera, Inc. bought the remaining assets of DR DOS in 1996. With this, Caldera was armed with the historical legal rights to the software. It used them to file a massive antitrust lawsuit against Microsoft in a Utah federal court. As the case moved toward a jury trial in the late 1990s, Caldera’s legal team gained access to Microsoft’s private corporate communications through pretrial discovery. The resulting internal emails completely shattered Microsoft’s “technical compatibility” defense. The ultimate smoking gun was a memo from Brad Silverberg — the senior executive who ran Microsoft’s Windows group and would soon ship Windows 95 — who explicitly outlined the psychological warfare behind the AARD code. Here’s a direct quote: “What the [user] is supposed to do is feel uncomfortable, and when he has bugs, suspect that the problem is dr-dos and then go out to buy ms-dos.” But that memo was only one slip in a long paper trail. The same discovery process surfaced a chain of emails from the autumn of 1991 (while Windows 3.1 was still in development) that read less like engineering notes. On September 30, 1991, Microsoft’s David Cole laid out the goal plainly: “It’s pretty clear we need to make sure Windows 3.1 only runs on top of MS DOS or an OEM version of it.” His colleague Phil Barrett spelled out the method the same day: “The approach we will take is to detect dr 6 and refuse to load.” “heh, heh, heh … my proposal is to have bambi refuse to run on this alien os. comments?” Barrett didn’t even bother dressing it up as a quality concern. In a separate note to Silverberg about the Bambi module, he wrote: “heh, heh, heh … my proposal is to have bambi refuse to run on this alien os. comments?” And in perhaps the most damning exchange of the entire case, Silverberg emailed colleagues that DR DOS “has problems running windows today, and I assume will have more problems in the future.” The reply came from Jim Allchin — a senior executive who would go on to lead the development of Windows XP and Windows Vista — and it was four words long, erasing any doubt about intent: “You should make sure it has problems in the future. :-)” “You should make sure it has problems in the future. :-)” The price of a rigged market 280 million to make the lawsuit disappear. While the settlement was an enormous financial victory for Caldera’s investors, it arrived far too late to salvage a competitive marketplace. By the turn of the millennium, Microsoft had welded its DOS and Windows lines together into unified operating systems like Windows 95 and Windows 98. The concept of buying a standalone, third-party DOS was completely obsolete. Microsoft had leveraged its ruthless tactics to secure a near 95% monopoly over the desktop computing world. The pioneer who never saw the reckoning Gary Kildall died in 1994 — two years before the lawsuit even began Gary Kildall never got to watch any of this unfold. On July 8, 1994, he walked into the Franklin Street Bar & Grill, a biker bar in Monterey, California, and suffered a fatal head injury. The exact circumstances were never settled and the autopsy couldn’t conclusively determine a cause. The timing is cruel. Kildall died in 1994. Caldera didn’t file its lawsuit until 1996. The incriminating emails didn’t surface in discovery until the late 1990s. The settlement landed in 2000, and the 280 million for being clumsy enough to write the intent down in an email. The lesson the industry took away wasn’t “don’t do it.” It was “don’t put it in writing.”