Where the Trial Was Not
We now enter the realm of application software to contemplate the impact of the Manifesto’s proposed reform. The existing software IP regime is valuable to both platform and application developers, but in subtly different ways.
The key difference emerges from a fundamental property of network economics—or at the very least, a property of business plans appropriate in network industries. If your product defines a network, you should always consider pricing it below its full value to maximize its circulation. If you can lock in your customers, you’ll find ways to reap significant profits later. If, on the other hand, your product sits on top of a network, you’re likely to have to sell it to profit.6 As a result, we can safely ignore the technical problem of classifying middleware, and simply use “platform” as shorthand for products that define networks and “applications” as shorthand for everything else. Though platform developers may consider many interesting business plans, application developers typically generate their most reliable profit streams via direct sales—a stream that’s only possible if they have an IP right protecting their object code.So we need IP rights to motivate innovation. But they knew that when they wrote the Constitution. Give them a copyright, and let’s move on. And we could do that. But since we’ve already seen one of the jeremiad’s prophecies come to pass, we should spend a little time thinking about just what software copyrights mean—particularly since something seems wrong. Copyrights last at least ninety-five years, which is effectively forever in the software world. Do we need to offer infinite protection to motivate application developers? And more broadly, could weaker rights motivate just as much software innovation?
Microsoft knocked out DR-DOS by launching Windows, receiving the standard package of IP rights, and leaving Digital Research to play catchup.
What could Digital Research have done? One option, at least in theory, was to clone Windows. But as Gates himself noted, Microsoft held some patents that would have made legal cloning tough; any software that performed the same tasks as Windows would likely have infringed. Digital Research’s other option should have been to get DR- DOS to talk to Windows, whether Microsoft wanted them chatting or not. Microsoft obviously agreed that Digital Research had the legal right to do so, because if it could have blocked DR-DOS from the Windows APIs just by stating that using those APIs was not permitted, it would have. Instead Microsoft blocked DR-DOS from chatting with Windows by keeping its source code secret. Microsoft figured out how to use its IP rights to extract stronger protection than anyone had intended to give it. And true its nature, Microsoft used those rights to knock out its primary competitor and to increase its own profits. Its doppelganger Musoft would have been proud.Speaking of Musoft, though, how did it handle such matters in Manifestoland? Well, when it prepared to launch Mundos—the first graphical interface to DOS—upon an unsuspecting but desirous public, it faced a choice. It could either reveal its Mundos source code, charge for each copy of its object code, and block competing products for a few years, or it could keep its source code secret, make its APIs a moving target, try to ensure that Mundos could sit atop only MU-DOS, and allow the Mundos object code to circulate free of charge.
Suppose first that Musoft had chosen the rights and revealed its source code in exchange for the short-term broad monopoly on offer in Manifestoland. Its chief competitor, DR, would have used its public access to the Mundos source code to bolster its own development efforts. In no time at all, DR would have developed new features and sought IP rights to protect them. That would set up a problem. Musoft couldn’t improve Mundos, because DR (and perhaps other companies) would own the rights to the next logical evolutionary steps.
But DR couldn’t sell anything until Mundos’ IP rights expired.Consumers might have been stuck paying a monopoly price for Mundos 1.0 and blocked from buying anything better—at least for a while. But once Musoft’s rights expired, the lot of consumers would have improved quickly because DR wasn’t the only innovative developer whose work benefited from access to the Mundos source code. The minute the Mundos rights expired, numerous competing second- generation products would have flooded the market, each compatible with Mundos 1.0, and each sporting its own set of protected advanced features. These competing second-generation products would have constrained each others’ prices. Third-generation systems, in turn, would have incorporated all desirable second-generation features—and each would be compatible with all second-generation systems. The process would continue from there. Each generation would get better and cheaper—but there could be occasional “dead times” between generations when consumers simply couldn’t buy the upgrades they wanted because Manifestoland’s broad IP rights blocked them from the market.
But such dead time is far from inevitable. Suppose that rather than blocking each other from the market, Musoft and DR had cross-licensed their rights and marketed Mundos 2.0 together. Consumers would have gained rapid access to the second-generation product, albeit at a monopoly price. Competition would then have moved on to the next set of fea- tures—which could also have resulted in cross licensing (or not). The bottom line, though, is that if Musoft had put the Mundos source code in the public domain, it would have had a very hard time knocking DR out of the platform market.
Suppose instead that Musoft kept the Mundos source code secret and simply let its object code circulate freely. That might have worked with Mundos because it was technically middleware, but it would prove disastrous with most applications. Musoft quickly would have discovered that it had no way to recoup its development costs—much less to turn a profit.
In other words, releasing an application unprotected by IP rights, even those on offer in Manifestoland, would have been inconsistent with Musoft’s nature. No, Musoft would have sought and received protection not only for Mundos, but for all applications that it developed in Manifestoland. Unless, that is, Musoft became an open-source advocate. But open-source licenses make little sense in Manifestoland, where all recipients of software IP rights open their source code. Open source advocates in Manifestoland can simply set their prices to zero and rely on consulting and customization services as their sole sources of revenue.Would the Manifesto’s edicts be good for consumers of application software? Prices likely would be higher, at least in the early stages. In exchange, consumers would gain some significant benefits, starting with intergenerational compatibility. Because everyone’s previous-generation source code for all protected applications is public, all developers would ensure that new products were compatible with all of their competitors’ old products. Furthermore, the fury of competing on each and every new feature would probably speed technological development. Intergenera- tional compatability and rapid product development would serve consumers well. The potential increase in compatibility is particularly significant given the network nature of the software industry. It seems inevitable that all important protected programs in Manifestoland would converge to open standards within their first few generations—a situation that can have significant positive consequences. Open standards lower entry barriers, increase competition, and eventually lead to lower prices.
But winners rarely exist in the absence of losers. If consumers have more choice, better products, and lower prices, and application developers like DR get to stay in business, only Musoft loses. But if Musoft loses by too much, we’d never see Mundos. Musoft, staying true to its nature, would decide that it can’t justify the investment necessary to develop a truly innovative new product that might never work—and even worse, that could be finished only a month after a competing new product received IP protection that blocked it from the market.
The Manifesto would have undermined the Constitution and failed to promote the progress of science and the useful arts. After all, worse things than monopolists who exploit consumers exist—and potential innovations that die on the vine are among them.The challenge is finding the right balance. If we’ve given Microsoft overprotective rights, we must be able to reduce them without reducing overall software innovation. That’s the definition of “overprotection.” In Manifestoland, we would have to make sure that the limited period of protection before we allowed competing products to enter the market, possibly as well as the lead time between the Mundos commercial launch and the release of its source code, were long enough for Musoft to reap the profits necessary to motivate innovative development efforts. How long is that? I don’t know for sure. Jeff Bezos reportedly suggested that three-year software patents should be sufficient. I’ve heard sillier suggestions—say 95 years. But however long it is, there should be some place in the general vicinity of Manifestoland where revised IP rights lead to both happier applications consumers and applications markets that are harder to monopolize and to exploit than are our own.
So much for where the trial was not. As we return to what we did at the trial, rather than watching the story as it unfolds, we’ll look beneath the surface to see how our real-world Microsoft was able to achieve things that would have eluded Musoft of Manifestoland.