Leader of the Pack

Behavior in battle is often controversial—and Microsoft’s has been no exception. In certain circles, it’s not at all uncommon to see the names “Bill Gates” and “Mephistopheles” linked.

Of course, not everyone shares that opinion, even within the techni­cal community. To many, Gates’s business model was pure genius. Con­sumer demand drove Microsoft’s entire product line. Whereas most engineers like to invent gadgets that excite other engineers, Gates looked for products struggling to earn a toehold in the consumer world, like the MITS Altair or the IBM PC, and supplied the missing piece. He scoured the market for newly launched ideas with potential and devised ways to turn them into popular products. He catapulted consumer computing decades ahead of where it might have been without him. Gates’s sup­porters are confident that Microsoft and its leader belong to Paradiso, not Inferno. A Google search shows Gates’s name linked far more often with God and Jesus than with Mephistopheles—and only a small frac­tion of these hits relate to the wonderful efforts of the Bill and Melinda Gates Foundation to improve the health of the world’s poorest people.⁶

In all likelihood, the truth lies somewhere in between (as it always does).

Liebowitz and Margolis, the skeptics who claimed the inherent supe­riority of the QWERTY keyboard, contend that inferior products never get locked in. They moved beyond typewriters to argue that VHS offered consumers a better combination of price and quality than Beta—chal­lenging another classic example of lock in. Then they turned to software. In Winners, Losers, and Microsoft,⁷ they claimed that Microsoft’s con­sistently superior products led to its many victories—and that superior competing products led to Microsoft’s few losses.

Build a better mousetrap and the world will beat a path to your door. This, surely, is one path to success. There’s nothing mysterious or underhanded about the success of a company that provides consumers with a product that gives more bang for the buck than its competitors But the possibility of lock-in suggests

that there may be other paths to market determination...... If and when there is

lock-in, a product succeeds in spite of inferior quality.... One interesting market to study is the software market, which is often alleged to exhibit network effects, lock-in, leveraging, and tipping Microsoft, by any reckoning, is a tremen­

dously successful company, but why is it so successful? Does it just build better mousetraps? Or, as some have claimed, are its products only mediocre? Has it achieved its large market share in spite of its mediocre products, by lock-in and luck, or through the leveraging of its ownership of the operating system? Our data provide clear answers to these questions. Good products win. Microsoft’s success derives from good business decisions and superior products.⁸

Their data analyses make a compelling case that Microsoft was more responsive to market demands than were its competitors, that Microsoft provided superior combinations of price and quality, and that the market rewarded those efforts by granting Microsoft market leadership—partic- ularly in the spreadsheet and word-processor markets whose data they studied.

Growing from just one of many competitors into a monopolist is tough. And that’s where Liebowitz and Margolis made a critical mistake. Their analysis of market data and product reports stressed the quality of Microsoft’s products, the savvy of its marketing department, and the wisdom of its business strategies. But even if Microsoft’s successful prod­ucts were far superior to those of its competitors, it’s hard to see how Microsoft could have monopolized these markets in the absence of network effects. Suppose, for example, that several consecutive genera­tions of Microsoft Excel were better than the concurrent releases of Lotus 1-2-3. It seems unlikely that the quality gap would have been sufficient to knock Lotus down to a niche without a bit of help. That help likely came from Excel’s working more smoothly with Windows. That edge represents a leveraging effect; Microsoft pushed from its monopoly in platforms to gain a second monopoly in spreadsheets (and a third in word processors). But even superior quality and leverage shouldn’t have been enough for Microsoft to monopolize a market—other than one that tipped to a standard. What’s more, Liebowitz and Margolis completely ignored the nature of lock in; though Microsoft’s products may have been unrivaled at the time that they locked in consumers, there’s no guar­antee that superior rivals wouldn’t emerge. Microsoft needed to create artificial barriers to stave off that possibility. Therein lay the antitrust violation and the harm to consumers.

But Liebowitz and Margolis were hardly alone in ignoring (or avoid­ing) this point. They were joined by the editorial pages of the Wall Street Journal—a consistent vociferous critic of the government’s case—and by the many callers to talk radio who asked: “Why can’t the government just leave Microsoft alone? They broke up the phone company and now look at the mess we’re in.” Questions of this sort suggest how poorly the scholars of both network economics and antitrust have educated the general public.

In a deeper sense, though, these talk-radio callers had correctly detected parallels to the telephone monopoly. Ma Bell had given us the best, most reliable, most comprehensive, least expensive communications system in the history of the planet. Who could have been blamed for thinking that it might not be a good idea to tamper with such a good thing? The same was true of Microsoft. Windows was—and remains— the best, least expensive, easiest to use, most widely accepted, most reli­able platform software standard the world has ever seen. Were we really sure that we wanted market cops tampering with a good thing like Windows—or with the valuable, admired company that developed it?

Yes. We were sure. Windows may have become the de facto standard because Microsoft’s remarkable knack for releasing the right product at the right time positioned it to have the most popular platform at the time that the market tipped to a standard. That quirk of timing arose in part through a combination of product quality, marketing savvy, and luck— the factors typically necessary to catapult a strong competitor to monop­oly status. But in the case of Windows, quality, savvy, and luck were not all Microsoft had going for it. Microsoft also had reached a strategic decision to skirt the borders of both IP law and antitrust law. And the genesis of that strategy takes us back to our historical detour.

By the mid-1980s, several competing versions of DOS were available; Microsoft’s MS-DOS was the most popular of them. Meanwhile, the overlords of IBM’s Big Blue Empire continued collaborating with Microsoft to develop the operating system of the future, OS/2, the plat­form that would simultaneously supplant DOS and cripple the Mac— thereby restoring the empire to the full reach of its glory. Microsoft’s little secret was that while it was intent on maintaining the empire’s integrity, it planned on staging a coup.

Between about 1985 and 1987, a couple of important new products hit the market. One was DR-DOS, first released by Digital Research in 1987. DR-DOS was fast, efficient, clean, and it allowed you to run any program originally written for MS-DOS. Digital Research, in fact, claimed that DR-DOS was superior to MS-DOS. Microsoft didn’t debate the point; its confidence in OS/2 had led it to more-or-less ignore MS- DOS’s development.⁹ DR-DOS thus threatened Microsoft’s bread-and- butter product. DR-DOS and MS-DOS presented identical translation frontiers. The two could talk to all of the same machines, they spoke the same API dialect, and they could converse fluently with all of the same applications.

Throughout the rest of the 1980s and on into the 90s, Microsoft faced real competition in the DOS market. At one point, DR-DOS captured ten percent of the U.S. market—and did even better overseas.¹⁰ And DR- DOS’s reputation for superior quality just grew. In August 1990, for example, BYTE magazine glowed that:

The latest incarnation of DR DOS, Digital Research’s MS-DOS clone, is an inno­vative and intriguing operating system that’s thoughtfully designed. Version 5.0 is also packed with the extra features that Microsoft’s own operating system should have (and might eventually have if the long-rumored MS-DOS 5.0 becomes a reality). As the people at DRI make very clear, its not pronounced Doctor DOS, although the analogy isn’t far off the mark, since it indeed cures many (but not all) of MS-DOS’s shortcomings.¹¹

Several months later, in its “Awards for Technical Excellence” issue, PC Magazine gushed:

Digital Research is the microcomputer operating system company that predates Microsoft. As if to prove it hasn’t lost its touch, DR DOS 5.0 does all the things you wish MS-DOS did.

Microsoft was hurting—and worried. Gates himself lamented that: “DOS being fairly cloned has had a dramatic impact on our pricing for DOS. I wonder if we would have it around 30-40% higher if it wasn’t cloned. I bet we would!”¹³

Microsoft fought back—on two fronts. The first lay somewhere in that netherworld between marketing and propaganda. Microsoft issued a number of strategic announcements about products and improvements that it was on the verge of releasing—products that might prove to be incompatible with DR-DOS. This practice, known in the industry as releasing “vaporware,” is an insidious way to frighten customers away from powerful competing products without really offering them an alternative.¹⁴

The second front was even more insidious. It involved that second important product launched in the mid-1980s—Microsoft Windows. Windows 1.0, launched in 1985, adhered to the gospel of modularity; it was a graphical interface that sat on top of DOS. DOS continued to define the translation frontier, and thus the network; Windows was an application that communicated with DOS. It was, of course, a special type of application, because in addition to talking directly to users (as most applications do), Windows also exposed its own set of APIs to com­municate with other applications. Such special applications are “mid­dleware” between the platform and the more conventional applications. True applications perform a task. They communicate with the platform at one end and the human user at the other. True platforms, on the other hand, embody full downward-translation chains. Middleware commu­nicates only with the platform at one end—it relies on the platform for the rest of the downward chain—yet with another application at the other end. The gospel of modularity implies that developers should launch potential platform innovations as middleware, where they’re easy to fix, to modify, and to improve without interfering with the plat­form. Each generation’s middleware programs provide candidates for migration downward in the next evolutionary stage of the platform’s development.

Microsoft set out to monopolize the DOS market by growing a network around its proprietary Windows APIs. Independent application develop­ers who wanted to take advantage of Windows’ graphical capabilities could write programs that spoke the Windows API language. Many developers chose not to go the Windows route. Programs that spoke directly to DOS tended to remain faster and more robust—if less attrac­tive and less user-friendly—than those that communicated with DOS only through Windows. Nevertheless, many other independent developers— as well as Microsoft’s own application team—did learn and use the Windows APIs. Many consumers appreciated this new combination of an inexpensive machine with the look and feel of a Mac. Windows-based applications came to represent a significant chunk of the market.

This burgeoning demand for Windows products within the DOS market posed a quandary for Digital Research. One of DR-DOS’s biggest selling points was that it could talk to all of the same programs as MS- DOS. In order to maintain that feature, DR-DOS needed to be able to talk to all of the Windows-based programs. Either Digital Research could write its own program that spoke the Windows API language, or DR- DOS could talk to Microsoft’s Windows. For obvious reasons, the first approach would have been more lucrative (after all, Digital Research might have picked up a significant chunk of the growing Windows market), but the second was easier—or at least, it should have been.

Microsoft ensured that neither could occur. Gates himself “doubt[ed Digital Research] will be able to clone Windows. It is very difficult to do technically, we have made it a moving target and we have some visual copyright and patent protection. I believe people underestimate the impact DR-DOS has had on us in terms of pricing.”¹⁵ This sort of secu­rity blanket combines the good and the bad. Microsoft’s ability to design and to develop a technically sophisticated product serves consumers well. It’s a perfect illustration of how IP rights motivate innovation; the Con­stitution would be proud. Microsoft’s decision to “make it a moving target,” on the other hand, serves no purpose other than to keep com­petitors out of the market. From the perspective of consumers, it’s a potential disaster. It diverts Microsoft’s time from product improvement to market protection. It prevents Digital Research (or anyone else) from designing a product that’s not only compatible with Windows, but that might even be superior to it. And, as Gates noted, it helps Microsoft keeps its price up. The net effect is that consumers in May 1989—when Gates wrote this e-mail—were paying more for a weaker version of Windows than had Microsoft been willing to work with only its IP rights and market forces.

In the face of these obstacles, Digital Research eventually fell out of the market. It sold DR-DOS to Novell, who eventually sold it to Caldera. In 1996, Caldera filed suit against Microsoft, alleging violations of the antitrust laws. Microsoft asked the court to dismiss the charges, but in late 1999 the court refused. In early 2000, Microsoft and Caldera reached an undisclosed settlement. That means that Caldera never proved its claims in court, and Microsoft can still legally claim that it did nothing inappropriate to wrong either DR-DOS or consumers. Along the way, though, a fairly sizable collection of the documents that Microsoft had had to turn over during the litigation made their way to the public; the e-mails quoted above were but a small sample.

So much for Microsoft’s skirting the antitrust laws. What of the IP laws? Here, despite arguably skirting the law, Microsoft did not violate it. The similarities between Windows’ graphics and those on the Mac desktop were hard to miss. They both used windows, icons, and mouse- driven point-and-click commands. Some of the icons even looked alike. Apple sued, claiming that the similarities were so close that they infringed its copyrights. It turns out that this type of claim is one of the toughest challenges for IP law to navigate. After all, if IP law allows me to take your product, tweak it in some minor way, and market it as my own, your IP rights aren’t worth very much. That type of rule would under­protect your rights and promote far too little innovation. If, on the other hand, my new product isn’t allowed to bear any similarity at all to any existing product, pretty much anything I do will infringe someone’s rights. That won’t motivate much innovation either. The framers of the IP clause would hang their heads in shame at either scheme. They would tell Congress to do better. But Congress would have to punt, because the question of “how similar is too similar?” isn’t really a question of either policy or law; it’s a question of fact. The courts need to consider each case independently.

Fortunately for the judges asked to look at these two complex pro­grams, others had already faced similar challenges. These judges found a test for software copyright infringement that had been floating around the courts for a couple of years, adapted it to the facts at hand, and com­pared the two platforms.¹⁶ They started with a list of all of the similar­ities. Then they divided their list into three categories: similarities that occurred because the design decisions were obvious; similarities that arose from Microsoft’s earlier relationship as a contractor to Apple; and similarities that arose because Microsoft infringed Apple’s IP rights. When they finished partitioning their list, they noticed that the first two categories were quite long. The third was empty. The courts ruled in Microsoft’s favor. And the final piece fell into place.¹⁷

Microsoft had launched Windows in 1985 into a competitive oligop­oly: a small number of large players paid attention to each other’s moves and attempted to steal each other’s customers. Hardware and software manufacturers both competed to add new features while keeping their prices down. Consumers choosing a system had to consider the manu­facturer’s reputation, the features it offered, and the price—with Apple providing a unique and somewhat extreme case of price/quality trade­offs. Software manufacturers also faced some difficult decisions about compatibility. To some extent, they all wanted their new systems to be “backward compatible” with those of their competitors, so that they could attract customers who used to favor their competitors. On the other hand, they did not like it when their competitors applied a similar strategy to win away their own customers.

Windows changed all that. Not immediately, of course; early versions of Windows weren’t very good products. But when the powerful Windows 3.0 appeared in early 1990, independent software developers helped Microsoft build a sizable collection of applications that spoke only the Windows API language. Microsoft had earned its first true appli­cations barrier to entry. IBM’s OS/2 was doomed from the starting gate. And the days of the Mac as a serious general-purpose rival were num­bered. Windows had thus achieved its two main objectives: it shrank the Mac’s share of the market and marginalized all competing successors to DOS. Between early 1991 and mid-1993, Microsoft’s stock doubled in price. IBM and Apple each lost about two-thirds of their market value.

Microsoft’s 1994 IP victory over Apple was indeed the last piece to fall in place. And with that, we finally understand how Microsoft became kingpin of the software industry. Quality, savvy, luck, timing, and a strategic decision to skirt the edges of the law all played a role. But what­ever the balance among these factors, their combination worked like a charm. Microsoft was finally free to turn to its next great challenges: the technical challenge of integrating Windows graphics and MS-DOS to create a single, smooth, next-generation platform to assume its proud role at the translation frontier; and the business challenge of maintain­ing the monopoly position that it had worked so hard to achieve.