The latest version of Windows, the operating system manufactured by the Microsoft Corporation, consists of some 30 million lines of tightly interconnected computer code: a programmer’s War and Peace, running to some 500,000 pages. Is that the right amount? Too little? Too much? These questions might conceivably intrigue certain specialists, like other programmers; or, possibly, connoisseurs of what David Gelernter has elegantly called “machine beauty”—the technological mating of simplicity and power; or economists charting market efficiencies. The last question—is it too much?—has perforce come to intrigue lawyers as well, and specifically antitrust lawyers charged with protecting economic competition and the interests of consumers.
But how is an antitrust lawyer to determine that a “manuscript” as long as Windows 2000 is economically dysfunctional—that it has been coded, packaged, or licensed to do consumers more harm than good? He cannot. Nobody can. But neither can he just throw up his hands in despair, on the grounds that, in this one arena alone, economic law can somehow be dispensed with altogether. This dilemma is at the heart of the mind-bogglingly intricate trial now moving forward under Judge Thomas Penfield Jackson. (Jackson’s “conclusions of law” may be issued this month, with “remedies” to follow thereafter.)
Things were simpler a century ago, at least as the history books tell it. Trustbusters did not need a lot of legal or economic theory to explain why they went after the Microsofts of that era. Runaway corporate growth was bad, and people just knew it. In those days, too, the obvious remedy for “bigness” was “breakup,” and that was how many of the landmark cases ended. A 1911 ruling by the Supreme Court divided Standard Oil into 30 separate companies. Judge Jackson knows this history well: “I don’t really see a distinction,” he remarked at a February 22 hearing, between John D. Rockefeller’s “control over his oil” and Microsoft’s dominance of the personal-computer (PC) operating system.
How big a legal difference there is remains to be seen, and may not be finally determined until after one or two rounds of appeals are exhausted. But there is certainly a political difference. The Clinton Justice Department and its assistant attorney general for antitrust, Joel Klein, are not just throwbacks to the populist trustbusters of a century ago, nor is the Microsoft litigation another grab for power on the part of a grabby administration. The truth is less political—and therefore less alarming—than that, if also, unfortunately, much less clear.
The basics of the Microsoft case are straightforward enough. In 1981, IBM asked Bill Gates to supply a “disk operating system” (DOS) for IBM’s first PC. That modest program provided the interface between the hardware—the screen, keyboard, disk drives, and microprocessor—and the software—mainly, “applications” programs for spreadsheets and word processing.
In the years that followed, Microsoft relentlessly added new features and capabilities. The operating system grew by over one million lines of code a year. Many of the new features had first been sold by other vendors as separately packaged applications or “utilities” for modems, memory management, CD-ROM’s, faxes, calculators, calendars, clocks, and cardfiles. But step by step, Microsoft inhaled them into DOS, and then into its successor, Windows.
With the expanding code has come an expanding array of contracts and copyrights, which define how the code may be used, repackaged, and resold. Software is extraordinarily malleable: independent vendors can (and do) graft on their own “dashboards,” “wallpaper,” and “skin”; they can (and do) change the inner workings for better or worse; they can enhance, infect, sabotage, or steal the code outright. Naturally, Microsoft’s marketing units and lawyers did all they could—and possibly an illegal amount more—to make sure the company retained control of what Microsoft’s programmers had crafted.
In 1994, a startup company called Netscape came up with an application to run on top of Windows. But this—the first good “browser”—was an application of a different sort. It supplied an interface not to the computer on the nearby desktop—Windows already did that—but to the countless remote computers that defined the World Wide Web.
Gates quickly recognized how serious a threat the browser-plus-Web presented to Microsoft’s core business. With the power of the Web above—power residing in countless mainframes and servers that use no Microsoft software at all—the browser threatened to displace much of what had been operating on the desktop below. Through the browser, the Web might well absorb most of the desktop—unless the desktop, at the very least, absorbed the browser first. Netscape saw it that way, too: Marc Andreesen, Netscape’s top technology officer, boldly promised to reduce Windows to an unimportant collection of “slightly buggy” “device drivers.”
Microsoft was not about to be reduced without a fight. It quickly created a browser of its own: Internet Explorer. It forged tight links among Explorer, Windows, and other Microsoft applications for word processing and spreadsheets, and equally tight contractual links with vendors of hardware and providers of online services who could promote Explorer over Netscape. And it made Explorer “free”—which is to say, Microsoft simply folded the price of the browser into the price of Windows. Gates did not want the market for browsers any more than Andreesen wanted the market for operating systems. (“Operating systems?” Andreesen asked rhetorically, “We don’t need no stinkin’ OSes!”) Both wanted to create an integrated platform that did it all.
Unfortunately for Andreesen, Gates had both the skill and the resources to win this battle in short order. But—so the government would allege—he used unacceptable tactics to do so. Last November, Judge Jackson agreed. Microsoft, he concluded, had unfairly jiggered the Windows code to favor Explorer over Netscape, and had bullied others in the industry into signing contracts to bundle Explorer, rather than Netscape, with their own hardware and online services.
There was no smoking gun in the trial, or in Judge Jackson’s voluminous first-round opinion. There was just a mountain of detail, an account of intemperate Microsoft e-mails, pugnacious negotiations, contractual fine print, refusals to return phone calls—line after line of detail, piled higher and higher, sort of like Windows itself. What was critical in the opinion was the definition of Microsoft as a corporate monopoly in one market—Intel-based PC operating systems—aggressively extending its line of business into an adjacent but distinct market—browsers.
Understanding the economic issues at stake here requires another narrative. It too begins almost a century ago, and it concerns the basic contours of corporate America, yesterday and today.
Built in 1917, the Ford Motor Company’s River Rouge plant in Dearborn, Michigan soon became one of the wonders of the industrial world, turning raw steel, raw rubber, sand, and silicon into automobiles. Ford made steel, made glass, made tires, assembled the parts—and did it all in one place. River Rouge absorbed the industries around it, or re-created them under its own roof.
But that is not the main trend today, at least not among what remains of yesterday’s corporate behemoths. In-sourcing has given way to out-sourcing. Industrial companies are being chopped up; economic life is being disintegrated into sockets and plugs—new, standardized units of production and profit, with one vendor being plugged into the next. Thanks in large part to digital technology, an automaker can now deal with an independent supplier of engines as easily as with an engine division in its own company. And when it is as easy to deal with an outsider, there is a good chance that the outsider will perform better as well. Electronic links allow the market efficiently to combine arm’s-length competition with extreme specialization.
That is why so many of yesterday’s giants are shrinking so quickly. Single companies that used to operate the many layers of an integrated operation now specialize in far fewer. The logical if somewhat fantastical limit might seem to be a multitude of one-worker companies, all interconnected by way of the Web. That would assuredly spell the end of the line for antitrust lawyers.
But not so fast. Even as the old corporation shrinks, the new corporation expands. Microsoft and Intel, for example, have both grown extravagantly, by adding layer upon layer to their flagship programs and chips. Monsanto was recently accused of trying to monopolize global markets for corn and soybean seed, and then food production worldwide—a farfetched idea, but not quite so outlandish as it would have been only a few years ago.
What is it that makes these new technologies so expansive? The basic story is set out in an insightful paper written in 1998 by one Timothy F. Bresnahan: “New Modes of Competition: Implications for the Future Structure of the Computer Industry.” Bresnahan, as it just so happens, is now the chief economist in the antitrust division of the Justice Department.
The old computer monopoly was IBM’s, and IBM did indeed operate much like the old Ford. It supplied everything: punch cards, processor, printer, software, service. By contrast, the new computer industry, like industry in general, is defined by plug-and-play layers: microprocessor (Intel’s Pentium chip), operating system (Windows), software application (Word), browser (Explorer), data link (modem + phone line, cable, or DSL), Internet service provider (Errol’s), “portal” (AOL), search engine (Yahoo), and higher levels of content beyond (Amazon).
Few doubt that the new industry structure is more competitive than the old, but, as Bresnahan notes, the competition in each layer does not seem to last. One company—an Intel, Microsoft, or Yahoo—comes to dominate each layer. And once established, dominant positions tend to endure.
Economists have a pretty good idea why information-centered businesses can readily expand; Bresnahan elaborates on why they endure.
First, information comprises a large share of the products and services involved, and information scales up at very little additional cost. The first Pentium chip or copy of Windows 2000 costs billions to create, but both can then be replicated at almost no cost at all. So a single vendor can often serve the whole market more cheaply than could two or more competing suppliers. (Indeed, Microsoft’s real challenge is not how to replicate Windows; it is how to stop others—like the good citizens of China—from doing so.)
Next, once a monopoly settles into place, “network effects” quickly stabilize it and lock in its advantages. Computer “platforms” are “social” networks—clusters of hardware and software provided by different vendors that operate harmoniously according to common but highly complex standards. Getting a lot of independent people to think and work in sync is very hard; getting them to stop, once they have finally worked out how to do so, is harder still.
The new monopolists also grow huge because their products are so adaptable—they thrive in many different markets. Intel’s microprocessors show up in PC’s but also in cars, toys, and microwave ovens. Oracle’s software supplies an almost universal platform for tracking all kinds of transactions, and such tracking lies at the heart of banks, credit-card companies, and most forms of marketing and retailing.
Finally, once established in the market, chips and databases and genetically altered seeds can be expanded from within, very fast. Intel has added gates to its chips as fast as Microsoft has added lines of code to its programs, doubling and redoubling their capabilities year after year. Amazon builds a first Web platform to retail books, then easily adds drugs, toys, and yard sales. Thus, Windows has grown like a metastasizing tumor, and so have the chips on which it runs, and so have the many layers of the Internet that it has helped create.
So far, these trends have been most in evidence in what we think of as “software” markets. But “software” has now become the main engine of productivity and innovation across the economy—in farming, manufacturing, financial services, every form of marketing and distribution, and entertainment. Billion-dollar rigs and tankers on the high seas are all choreographed now by gargantuan computer programs that determine where to drill, how to pump oil from rock, how to move it, and how to price it. Hotels, air travel, the building of jets, and the manufacturing and selling of cars remain material enterprises around the edges, but most of the value and profit now depends on the information systems that design, route, book, and track.
In this manner are set the various layers of an industry slouching toward monopoly. If Exxon’s software for spotting oil two miles below an ice shelf is just 20 percent better (say) than everybody else’s, it is economically efficient and perhaps inevitable for Exxon to displace all other contenders in the drilling end of the business. The same goes for some other company (Shell, let us say) in the next layer if its computers give it a decisive edge in the very complex business of refining crude oil into gasoline. And so on, up the stack.
How the monopolist in each layer emerges does not much matter, at least not after the fact. It may emerge because it was the best during the critical early years when the Manichean competitive battles are fought and won. Or it may emerge by historical accident, or dumb luck, with or without the help of nefarious scheming and bullying. But once a platform is widely established, it actually becomes “the best,” in economic terms, however inferior it may “really be” in the eyes of people (like aficionados of Apple’s operating system) who think they know better.
Equally clear, however, is that once a platform monopolist gets established, it can expand its turf pretty relentlessly, whether to improve the product or to crush potential competitors. And money is no object: after all, at stake for Microsoft in any challenge to Windows is the sum total of all future profits from that stupendously valuable franchise.
The scenario Bresnahan foresees is what he calls, borrowing a phrase from the evolutionary biologist Stephen Jay Gould, “punctuated equilibrium.” For extended periods of time, platforms will simply improve, layer upon layer, and grow more entrenched as they do. But eras of stability “will occasionally be punctuated by epochs of radical change.”
In the world of stacked monopolies, such transitions do not have to happen often to keep markets efficient. Each layer exists in a meta-stable state, perched between stable cooperation and what may abruptly become terminally lethal competition. That is why near-monopolist Bill Gates can testify, in all sincerity, that competitors surround him. That is why his company has behaved, all along, as if they really did, and why it was determined to kill Netscape. For we now know—in retrospect—that the advent of the browser did seem to present the possibility of a serious “punctuation” in the industry. Gates himself saw it that way, mobilized his legions, and fought back as hard as he could—as every good and honest competitor should.
So what should antitrust lawyers do about all this? That Microsoft won the browser battle tells them nothing useful: somebody had to win. Still less can antitrust law turn on economic definitions, including the vigorous assertion, accepted by Judge Jackson, that “operating system” and “browser” are “separate products,” and that therefore Microsoft’s conduct can clearly be said to have engaged the old-style antitrust law of “tying” (bundling disparate products), “market foreclosure” (signing sweetheart deals with key suppliers or customers), “predatory practices” (cutting prices too aggressively), or “leveraging” (extending monopoly lines of business into adjacent markets). Just as no economic definition or theory is going to tell us whether an operating system should stop at 50,000 lines of code, or at 50 million, there is no objective way to ascertain from the inner workings of the code whether folding in new features—like a browser—represents the “tying” of two products or the “improving” of one, still less whether a two-box sale is economically more or less efficient than a one-box alternative.
Nor is monopoly itself a fitting target for today’s antitrust law. Intel, Windows, Exxon, Amazon: as Bresnahan admits, one company is quite likely to emerge in each layer, with the software and social network that it takes to dominate, if not monopolize outright. That is just the new nature of things. Moreover, in this same new nature of things, it takes a monopoly to fight a monopoly; in order to stand a chance of competing, “a firm absolutely needs a reasonably defensible ‘home monopoly.’ ” This, from the man who is currently the chief economist at the Department of Justice.
And so, if the fate of the old antitrust law was to protect competitors from each other that they might live to fight another day, the fate of the new, it seems, is to protect monopolists from each other—that they may live to fight another day. To accomplish this, the law has to focus, paradoxically, not on the big things but on the little things: those intemperate e-mails, pugnacious negotiations, and overreaching contracts. And perhaps—as Harvard’s Lawrence Lessig argues in his new book, Code and Other Laws of Cyberspace—on some of the equally fine details of the underlying computer code. Mountains of detail will supply the best evidence we are going to get. Motives will be much of the story; sometimes, all of it.
It makes us uneasy when huge lawsuits hinge on mountains of small, often subjective, detail; but, let’s face it, law often does. First-amendment and equal-protection jurisprudence, for example, frequently pivot on legislative intent, if only because scarcely a law can be written that will not somehow, somewhere, have an uneven impact, even if (usually) unintended. And there is something to be said for the notion that law and order are to be maintained by policing the little derelictions. An analogy here comes, ironically enough, from the “broken-windows” theory set out by James Q. Wilson and George L. Kelling in a famous 1982 essay. Police the little things—the outer skin, the thin interfaces—and the big things will quite often take care of themselves.
Like it or not, that is where antitrust law is headed. Civility in the software neighborhood will be maintained not from the top down, but from the bottom up. It will be a matter of attacking blight, in code and in the contracts that go with it. Antitrust prosecutors will fight the symptoms of disease, because they cannot objectively define the disease at all.
And how will they remedy those symptoms? Timothy Bresnahan asked that question in the article I have been citing. In answering it, he offered the image of a dog chasing a fire truck:
They, Microsoft, are the fire truck. We also have the prosecutor and political supporters salivating at the prospect of a chase. Looks good. Right down the middle of the street, loud, red, crowd already assembled. But gentlemen, what are you going to do with it when you catch it?
As with everything else in this story, simple answers are maddeningly elusive. One proposal is to clone Microsoft: break it in two, give each new half of the company a complete copy of the Windows code, and set them loose. That would mean competition; it would also mean, in the short run, the demise of the Windows standard. And in the long run, it would not last—not if Bresnahan’s theory is right. Monopoly will reemerge in the Windows-Intel layer of the market because economies of scale and network effects make it inevitable.
A second proposal is to break the company vertically, creating one owner of Windows and one owner (or maybe two) of Microsoft applications like Office and Explorer, and Microsoft’s online services (MSN). If the antitrust authorities believe their own theories, they would have to leave these several entities free to attack each other—vertical competition being the only real hope for competition of any kind in these markets. But if owning the operating-system monopoly really does still confer an unfair competitive advantage, as the antitrust prosecutors have asserted, we should then expect, again, an inexorable march back toward a Windows package with a bundled browser.
However enticing all this may sound to legal academics, none of it can sound good to the two classes—consumers and competitors—who are said to have been Microsoft’s principal victims. Few consumers would welcome such disruption: most of us, at this point, would surely rather stick with Microsoft and let it get on with the business of improving what we all now rely on, and we can hardly be cheered by the prospect of an assistant attorney general coming in with guns blazing to set us “free.” Nor would most of Microsoft’s main competitors feel any different. AOL—which now owns Netscape—knows that its future depends on the rapid continuing growth of well-equipped digital consumers, and it has far more to lose from the collapse of the Microsoft platform than from its robust survival. AOL’s own business is thriving several layers above that very platform.
With no real support anywhere for a breakup, I doubt it will come to that in the end. (I have suggested elsewhere that Bill Gates should take the initiative and orchestrate his own breakup while he can still shape the terms, but imposing such a remedy on him would be quite another matter.) But if not a breakup, and if not a remedy involving money—the government has not brought that kind of case—there will have to be some kind of injunction, a “future conduct” remedy. According to its terms, Microsoft would be ordered to eschew certain specified strong-arm tactics, or quarantined from entering certain markets, or directed to publish some key parts of its code or maybe (at most) to spin off a division.
The main trouble with such “conduct” remedies is that they often lead to years or decades of stifling judicial supervision. They also tend to stifle competition along the fault lines being policed, which is to say, precisely where competition remains the last best hope for competition of any kind. Bottle up the monopolist to the south of the divide, and you foreclose even the potential for competition against the monopolist to the north.
A final irony is that whatever form the remedy takes, it may turn out to have been historically (as opposed to legally) unnecessary. The future the Justice Department was anxious to protect by joining the case on the side of Netscape has already been assured. The Windows-Intel desktop has peaked, and everyone from Silicon Valley to Microsoft headquarters in Redmond, Washington knows it. Henceforth, exactly as Netscape’s Andreesen foresaw at the outset, capabilities will migrate out to remote servers and the Web, and users at home and in offices alike will be running their “net appliances” and “personal digital assistants” on stripped-down, non-Windows operating systems. By the time the Supreme Court ever gets to see this case, Microsoft’s hegemony will be history.
Does this complete fizzle of a conclusion mean the case should never have been brought at all? Part of me longs to answer yes, but the serious reply must be no. The case has already had a direct impact on Microsoft’s practices. The company has moderated its strong-arm tactics. Contracts have already been redrafted and toned down. Tight links between Windows and browser code have been severed. Those are all defensible results. It can reasonably be maintained that, going forward, they will enhance the overall level of competition in the industry.
Other companies with dominant positions in one layer of the market will certainly hesitate, in the future, to craft codes or contracts in the ways that got Microsoft in trouble. At Senate hearings in late February, the chief executives of AOL and Time Warner promised to open up a key interface—between Time Warner’s cable properties and AOL’s online services—that, until recently, Time Warner had been very determined to keep closed. This, too, from the point of view of the public good, is a defensible result.
Not long ago, the Bell System’s phone switches were programmed to hand off all long-distance traffic to a single carrier, AT&T itself, not to MCI or Sprint; and Bell’s network, and its customer tariffs, were designed to work only with Bell’s own phones, not Radio Shack’s. The most important and controversial sections of the massive 1996 Telecom Act, and the FCC regulations implementing it, concern the “unbundling” of the local phone company’s “network elements” right down to the level of virtual space (bandwidth) within the individual telephone line leading to your home. The electric-power industry is now in the throes of being unbundled along similar lines, with generation being untangled from transmission and the different layers of transmission being uncoupled from each other.
What lies ahead is more of the same. For “operating system” and “browser,” substitute any number of other hardware/software or software/software pairs that define the layers and interfaces of the new digital economy: “cable programming” (like CNN) and “distribution channel,” for example; or “Visa/Mastercard network” and “competing bank”; or “airline reservation system” and “airline.” This seems to be the way that software-dominated industries evolve—and, to a significant extent, we are all software industries now.
Last year, I suggested in these pages that litigation was precisely the wrong way to decide how to regulate our favorite poisons (“Guns, Tobacco, Big Macs—and the Courts,” June 1999). But those poisonous products are static. We know exactly what they do, and how they are used and abused. We have debated their legitimacy for decades, in both legislative and regulatory arenas, both federal and state. Litigation cannot improve the quality of this kind of public debate, it can only detract from it.
Software and digital hardware, by contrast, evolve so fast that they remain forever new. We do not have a Bureau of Operating Systems and Browsers. And the bored-to-tears factor ensures that we will never have a really good public debate about the interface between programs and applications. Dispersed, case-by-case resolution of clearly focused controversies is the worst possible way to make law here—except for the alternatives.
Case by painful case, a common law of software code will thus emerge from the decisional mists, through the gradual and erratic accumulation of complaints filed, settlements agreed to, judgments entered, and appellate opinions rendered. If Judge Jackson gets it wrong, other judges deciding other cases can get it right, or at least less wrong. And however wrong the judges may end up getting things, they cannot get them as wrong as a Federal Computer Commission could: wrong for the whole industry, all at once, a wrong meticulously spelled out in hundreds of pages of the Federal Register.
This does not mean that such cases should be easy for the antitrust authorities to win. On the contrary: it is hopeless, and almost certainly harmful, to fight every market-expanding step taken by Intel, Microsoft, Amazon, or Exxon. The best programs are likely to take over huge market shares, and should. The most the antitrust police can hope to do is to spot the arrival of exclamation-point changes in technology, and then see to it that established monopolists do not go overboard trying to secure the status quo. It is for the rest of us to police the police.