Personal computers are an example of what economists Timothy Bresnahan and Manuel Trajtenberg call a “general purpose technology (GPT)” in that computers provide a platform on which many other technologies can be built. Electric motors, steam engines, and radios are other examples of GPTs.

At an early stage of development, GPTs can be easily moved between applications: a particular gasoline engine might be used in a horseless carriage, a tractor, or even an airplane. This flexibility allows innovators to apply new technologies in a variety of unforeseen applications. But, as time goes on, the general purpose technologies become more specialized: one would never think about putting an automobile engine into an airplane today.

So in one sense, it is not surprising that a GPT like a personal computer would evolve to occupy specialized niches. Two common examples are cash registers and ATMs, which at heart, are composed of a standard PC motherboard and operating system that are then linked to a specialized user interface. Today, your smallest mom and pop market can have cash register technology that twenty years ago only the largest retail chains could afford.

A standardized GPT can serve as a base for future innovation. For example, if I need an electric motor for my latest gizmo, I can buy a standardized product out of a catalog. Similarly, if I need a database for an application, I can choose one from several alternatives.

I think Jonathan Zittrain would agree that this sort of evolution towards standards is generally a good thing. What he worries about (correctly) is that the “general purpose” aspect of the general purpose technology may be lost.

The danger he describes is that a general purpose device can be used to do harm as well as to do good. The same gasoline engine that powers a tractor can power a tank. The same PC used to play innocent games on the Internet can be reprogrammed as a zombie to initiate denial of service attacks on other computers.

Given that a general purpose device such as a PC can be reprogrammed in the service of mischief and mayhem, Steve Jobs is right to worry about selling a cell phone that is too open. It is not hard to imagine a scenario in which an open-source cellphone is infected by viruses that turn it into a zombie, capable of causing significant harm. Imagine what would happen if thousands of cell phones were to call 911 simultaneously.

Zittrain’s Dilemma is how to gain the benefits of an open, general purpose technology—such as a programmable networked PC—while avoiding the costs—malware that can infect an entire network of devices.

He offers several sensible suggestions for policy reforms to address this sort of problem. But there is no silver bullet.

Consider, for example, his discussion of the “norm of protecting settled expectations.” He proposes the following principle: nothing should prevent Microsoft keeping the Xbox closed, but if Microsoft opens it to third-party developers, it should not later be allowed to impose barriers on the operation of that third-party software.

Though one can applaud the sentiment, such a policy would be tough to implement in practice. After all, every release of Microsoft Windows breaks some applications—even Microsoft applications. Such failures may be legitimate; for example, a developer could exploit an undocumented feature that is later dropped. Some sorts of bit rot—the creeping obsolescence experienced by older software as operating evironments change—are more insidious: perhaps a third-party game runs more slowly because it can’t take advantage of an undocumented feature in a new version of Windows that only Microsoft knows about.

Ultimately, the best protection against bait and switch is an informed (and even paranoid) buyer who demands openness as a condition of purchase. One benefit of Microsoft’s “embrace, extend, extinguish” strategy is that it has made it very hard for new companies to play that same game. Nowadays, buyers demand data portability, interoperability, and flexibility at time of purchase, making it difficult for a seller to lock users in to a proprietary technology.

Consumer distrust of constraint reflects the nature of the Internet as “a lab experiment that got loose.” I have often described the Internet this way because one would never expect to see such an open and extensible technology like the Internet released as a business product. Think of the difference between the Internet and cable TV or a cell phone network. That said, it is safe to say that the Internet could never have developed as it has without the underlying infrastructure provided by profit-seeking telephone and cable networks.

So there will always be an uneasy peace, with occasional outbreaks of war, between closed, proprietary models and open ones. The challenge is to maintain sufficient diversity to allow each model to succeed in its proper place.