Get our latest essays, archival selections, reading lists, and exclusive content delivered straight to your inbox.
Internet Architecture and Innovation
Barbara van Schewick
MIT Press, $45 (cloth)
The Master Switch: The Rise and Fall of Information Empires
Knopf, $27.95 (cloth)
In 2003 Tim Wu, a professor at Columbia Law School, published an article on the once-sleepy subject of telecommunications policy. In it, he coined the term “net neutrality” to capture the idea that network operators—the Comcasts and Verizons of the world—should not be in the business of regulating the information traffic that passes through their networks. The term took hold, and the article launched Wu to cyber-rock-star status.
Net neutrality is a simple idea with powerful implications. A neutral net would, for example, prevent cable providers from slowing down their customers’ connections or, worse, banning them from running certain services. That is good for customers, who get equal treatment whether they are streaming movies on Netflix, chatting on Skype, or shopping on Amazon. And it is also good for Netflix, Skype, and other companies that have grown using an Internet infrastructure they do not own and have been able to innovate without worrying about shifting rules of the road.
With those credentials, net neutrality seems like a winning policy. But what about the network operators? They are not so happy with net neutrality, and it is easy to see why. If they respect net neutrality, they cannot impose special burdens on consumers who occupy lots of bandwidth by running data-intensive applications during periods of peak use. Nor can they ban Internet services that compete with their own offerings of cable TV or telephony, thus denying them a lucrative source of revenue. The result may be an underinvestment in infrastructure improvement, which is not good for Netflix and Skype, which depend on fast and ever-improving networks. Predictably, then, network operators prefer that the government not tell them how to run their networks and embrace industry self-regulation instead.
Despite this potential conflict between service and network providers, things have worked out pretty well so far. So even as battles over net neutrality have heated up over the past year—with important rulings from the Washington, D.C. Federal Court of Appeals and the FCC, and Verizon’s appeal of new FCC regulations—opponents often ask: why pass laws or regulations ensuring neutrality if we are doing all right without them?
In his new book The Master Switch, Wu contends that we need net neutrality now more than ever. His argument turns on an account of the history of previous information technologies. The nature of information as a commodity leads, according to Wu, to “information empires”: firms that dominate the industry, marginalize or destroy all competition, and stifle disruptive innovations. Thus AT&T suppressed the development of magnetic tape and with it the answering machine; RCA delayed the takeoff of FM radio; network television broadcasters tried to undermine cable.
Most of these companies started as benign and smallish innovators only to end up as giant corporations single-mindedly committed to preserving their market dominance. They also built vertically integrated businesses that, while extremely profitable, were socially harmful. If AT&T hadn’t banned its customers from connecting gadgets to its network, the Internet could have arrived much earlier. But AT&T controlled the networks, offered telephone services, and sold the phone sets. It had no interest in third-party innovations.
The open network and unbridled innovation of the last two decades have only been possible, then, because the Internet has been in an early phase in its life story. Today’s Internet innovators have been lucky because real information emperors have been busy elsewhere—fighting wars over the future of print or television. Now that there are many more eyeballs and opportunities online, the situation will change drastically.
To head off the emerging threats to openness and thus to innovation, Wu proposes that all firms in the information industry adhere to what he calls the “separations principle,” which would establish “a salutary distance between each of the major functions or layers in the information economy.” Broadly speaking, Wu wants to isolate three parties from each other: those who develop information, those who develop network infrastructure through which information travels, and those who control how information is accessed.
Wu’s other proposal, which he also folds into the separations principle, is less novel and radical. He wants to apply to the Internet the legal concept of the “common carrier,” which prohibits discrimination in sectors essential to the public good (e.g., transportation). This would prevent network operators from engaging in discriminatory and exclusionary practices, such as prohibiting access to Skype or Netflix.
Wu’s argument is hard to evaluate because it is difficult to be sure about the scope of his ambition. Readers who are familiar with his work on net neutrality are likely to interpret the historical examples discussed in The Master Switch in the context of that particular fight, thus as a defense of net neutrality. Read in this less ambitious way, the book is largely successful, though puzzlingly indirect.
But one could also read The Master Switch as a much bolder attempt to influence the future of the information economy, not just net neutrality. In the book and in recent public appearances, Wu has focused on the growing power of Apple, Facebook, and Twitter—not the usual contestants in net neutrality debates. He believes that some of these companies exhibit features of earlier information empires and may be hurting innovation. The separations principle is clearly meant to apply to them, and to the information industry as a whole, not just to network operators such as Comcast and Verizon. The merits and implications of Wu’s position, therefore, need to be assessed in a much broader context than net neutrality alone.
Implementing the separations principle could reshape much of today’s information economy. Establishing a clear dividing line between content producers, content disseminators, and those who control points of access means that Apple won’t be able to sell music and video via iTunes; Amazon to manufacture Kindles and publish books; Google to display search results to its own non-search services in its own browser.
Is this a good idea? The only contemporary example that Wu provides in support of the separations principle is the acquisition of NBC (a content-producer) by Comcast (a network operator). Wu’s concerns about this deal are probably justified: with all those NBC programs to peddle, Comcast may be hostile to Netflix and other services that distribute non-NBC video. But does such a deal warrant introducing a policy that might hamper innovation?
What if Apple had been subject to the separations principle when it entered the music market? When it launched the iPod, there were plenty of other mp3 players around. Would Apple have gone down this track if it had not expected to reap extra profits from selling music, mixing a technological advance with a novel business model?
Wu devotes a significant chunk of the book to Apple, but he doesn’t make a convincing argument that the company poses a threat that would justify subjecting it to the separations principle. More likely, Wu’s dislike for Apple is fueled by anger at the company’s decision to assert control over both its hardware and software. But Apple’s closed architecture is not necessarily harmful just because it is closed. In the absence of real evidence against Apple, Wu relies on a highly stylized historical account: Apple may not cause any harm at the moment, but Steve Jobs exhibits all the traits of an information emperor, thus Apple is an information empire; all information empires eventually become socially harmful, so we’d better act before it is too late. I am guessing that the ascension of Tim Cook in place of the ailing Jobs will not give Wu pause.
Highlighting Apple is an odd choice in a book that primarily deals with information networks. Apple does not, after all, install cables or mobile towers. Rather, it is in the business of producing slick gadgets and selling third-party content (music, movies, ten billion app downloads) to bring those gadgets alive. Wu never draws much of a distinction between Internet infrastructure and the broader Internet ecology that has emerged on top of that infrastructure. But the infrastructure and the ecology are subject to completely different dynamics. For instance, many of us live in neighborhoods with only a single broadband provider, but this is rarely the case with search engines or mobile phones.
Is Apple’s restrictive App Store policy analogous to AT&T’s erstwhile ban on external appliances on its network? Wu sees in both a centralized and harmful approach to innovation that also limits free speech. Apple’s short-lived prohibition of an app designed by a Pulitzer Prize–winning cartoonist is one example.
But the cases of Apple and AT&T are fundamentally different. AT&T controlled the only network available to the device innovators whose early technologies could have put the Internet in people’s homes years before that ultimately happened. Apple’s App Store, on the other hand, has to compete with Google’s Android Market, Nokia’s Ovi Store, BlackBerry’s App World, and many others—not to mention unaffiliated programmers. Anyone who hates Apple’s policy on free speech could leave and start creating apps for another platform; many developers already have. Apple’s centralized approach to innovation becomes problematic only if Apple is the sole supplier of mobile phones or mp3s or tablets: not impossible, but hardly on the horizon.
Wu never does justice to Apple’s side of the innovation argument. Apple sees immense benefits in centralization and control, in integration rather than fragmentation. Compare Apple’s approach to the app economy to Google’s: Google chose to let thousands of apps and app stores bloom and deal with the consequences later. That is a smart strategy for the underdog, but of uncertain benefits in terms of innovation.
Google’s decentralized process can be burdensome. The company has little control over the hardware that runs Android, which means the software now functions on extremely slow gadgets that can’t run applications written for higher-end models. In order to produce an app that works on all those Android phones, developers have to meet a set of requirements far more diverse and arduous than they face in building software for all iPhones. As the founder of an Android-focused application-development company complained in November 2009: “Instead of working on updates to our apps, we find we are trying to make each app work for multiple versions of the OS and different hardware capabilities . . . . if you are a small or a new Android developer coming in and trying to learn I could see your head exploding.”
The tightly controlled iPhone does not have that problem. Apple also need not worry about handset manufacturers or carriers who want to modify the operating system or install something proprietary on it. Is the risk of not identifying the next Skype greater than the benefits that come with centralization? It is hard to tell, but Wu does not even acknowledge what is ultimately a cost-benefit calculation, for he rejects Apple’s approach on ideological rather than economic grounds. For Wu, Apple’s closed architecture is contrary to the open spirit of the Internet. And maybe it is, but that is a shaky foundation for regulating Apple out of existence or into a shell of its former self.
Similarly, it will not do simply to assume that Apple’s control over the content in its App Store threatens freedom of expression. Sure, Apple has made plenty of bad decisions on this front, but in assessing the company’s contributions, positive and negative, to free expression, one should also consider the benefits the iPad might bring to newspaper and magazine publishing. There are already two major publications produced exclusively for the iPad, Project, a magazine, and The Daily, a newspaper. And many print outlets are looking at Apple’s closed system as a potential savior, a way to monetize content they were forced to give away on the Web.
In short, the contributions of modern information empires, whether Apple or Facebook or Google, need to be judged on a much broader set of criteria than those Wu outlines. Wu’s sweeping claims about the future of the Internet as seen through the prism of information empires and the alleged need for separation miss the mark. This is too narrow an analytical basis for thinking about Internet policy as a whole.
But could the information empire still be a useful concept in the net neutrality debate? Internet Architecture and Innovation, a new book by Stanford Law professor Barbara Van Schewick, shows that it is not. Reading Van Schewick’s book after Wu’s is like walking into a three-hour academic lecture after watching an eighteen-minute TED talk: while the lecture might seem too demanding at the outset, it eventually proves far more rewarding.
Internet Architecture and Innovation explores how changes in a system’s technical architecture and design affect the environment for innovation. Van Schewick argues that modularity—the degree to which a system’s components can be designed, produced, and used independently of one another—is a defining feature of systems. Lego blocks and personal computers score high on modularity; automobiles, less so. In highly modular systems, designers of a particular module can treat all other modules as black boxes; the design of their own module is independent of how other modules are designed or connected. In general, high modularity is good for innovation, as it provides designers with flexibility while also narrowing their focus to a single problem.
“Layered architecture” is one form of modularity. With layered architecture, all modules are arranged hierarchically; the modules assigned to a particular layer are only allowed to use the modules that lie on the same level or below it, and lower layers are shielded from changes to those above. Such designs are all around us: we do not typically need to change the foundation of a house in order to change the roof. This configuration makes a system less complex and easier to modify.
A similar layering principle is at work in Internet architecture, but with a twist. As a network, the Internet functions with two kinds of computers: the computers (like the one I am using to write this article) that use the services of the network to communicate with one another at the edges of the network, and the computers that help to implement the network (the servers and other intermediaries that lie at the core of the network and help to transmit data to and from its edges).
A service such as Skype can be offered in two ways on such a network. One is to enable both types of computers—those at the edges and those at the core—to recognize Skype data and make those computers handle that data with special care, different from the way they handle other packets of information. The other strategy is to keep the computers in the core of the network in the dark about Skype and simply have them pass packets of data from source to destination, without being able to recognize the contents of that data.
If the only service that we need to worry about is Skype, the solution is straightforward: it’s best to teach core computers about Skype, if only to ensure that no data get lost in the transmission process. But what if a network might need to handle millions of services, all of them demanding slightly different treatment?
This requirement is best served by a network with minimal complexity and specialization at the core and greater specialization at the edges. This idea—the “end-to-end” principle—guided the early development of the Internet. The principle’s broad version postulates:
A function or service should be carried out within a network layer only if it is needed by all clients of that layer, and it can be completely implemented in that layer.
In practice, the broad version of end-to-end makes it impossible to block services such as Skype, for this can only be achieved if the network operator starts manipulating data packets passing through the network. But this would require implementing an additional network function that is not needed at that particular layer, thus violating the end-to-end principle.
Van Schewick argues that the pioneers of the Internet recognized the network’s revolutionary potential, so instead of optimizing for performance or cost at such an early stage, they decided to maximize long-term evolvability. And that meant keeping the network core as simple and unspecialized as possible.
It is possible that the trade-off made by computer scientists in the 1980s may no longer reflect the needs of today. “If we believe that all important applications have been realized,” Van Schewick writes, advocating on behalf of the devil, “there is no need to incur the costs of keeping the Internet open for new applications.” But given that there was no Kindle, iPad, or Twitter just five years ago, that belief is a foolish one. As Van Schewick puts it, “Leaving the evolution of the network to network providers will significantly reduce the Internet’s value to society.” In other words, a non-discriminatory—neutral—Internet is in all our interests.
Most provocatively, Van Schewick argues that even the presence of competition may not succeed in thwarting discriminatory practices. And here is where she undermines Wu’s claims about information empires.
In most industries, discriminatory behavior by a firm in a competitive market gets penalized by rivals who offer the excluded goods to disappointed customers. Why doesn’t this happen in the market for Internet services?
First, the initial act of discrimination by one network operator may strike such a severe blow to the providers of excluded goods that they would exit the market altogether. If Comcast or Verizon blocks access to Netflix, the ban might not kill the company immediately but it could contribute to its eventual downfall, even if others don’t block it.
Second, competing network operators may choose to block the same application, leaving users nowhere to turn. It seemed like Neelie Kroes—who is in charge of telecom policy for the European Commission—was on to something when she urged Europeans to vote with their feet and leave mobile providers that restrict Skype. But this is not an option in France, where, despite strong competition among providers, all of them restrict Skype.
Third, network providers may hide their discriminatory practices, slowing down rather than blocking the applications they don’t like. Since consumers don’t know whom to blame for their slow speeds—it could be a problem with the application, not the network—they may stick with their providers anyway.
If Van Schewick is right, many network operators will discriminate even if they aren’t Wu’s information empires: the focus on information empires is a distraction, even for the analysis of net neutrality.
If network providers discriminate, the results could be damaging for innovation. Innovators who know that discrimination is possible or likely will expect lower profits from realizing their innovation and thus tend to underinvest. Moreover, the uncertain competitive environment might also scare potential innovators. And a network operator might imitate those services it allows to succeed and then exclude the third-party originator of the service from the network.
But how do we encourage network operators to continue expanding and improving their networks if increasing demand for bandwidth from high-traffic services undercuts the operators’ profits? Van Schewick recognizes that this could pose a serious problem. “If we truly believe that network providers need more profits to motivate them to deploy broadband networks,” she writes, “tax cuts on broadband investments or direct subsidies may be an option.” Another solution she proposes is public provision of infrastructure.
Sadly, this is all she says on the subject. We are left to wonder how policymakers should go about evaluating the claims made by network operators and how public rather than private provision of infrastructure would affect the innovation dynamics of application and content providers. To be sure, these questions are not the focus of her book. Still, this principal objection to net neutrality—the challenge of innovation at the level of the network itself—deserves more than a few sentences in the book’s last chapter.
In addition to effects on the future of the U.S. Internet industry, the outcome of the net neutrality debate has geopolitical ramifications that Van Schewick, Wu, and just about every other participant ignores. Discriminating between different types of content on networks requires monitoring that content, so the network operator would have to deploy software and hardware tools that reveal the kinds of data passing through its system. Comcast may only wish to throttle peer-to-peer file-sharing services, but many of the tools it would need could aid authoritarian states in censoring political speech and spying on dissidents. Iranian authorities are already moving in this direction, with their tight embrace of “deep packet inspection” technology.
Omissions aside, Van Schewick’s book makes a key insight on the persistence of incentives to discriminate even in a competitive environment for network connectivity. She also highlights the troubling aspects of Wu’s prescriptions. He is right on the net neutrality issue: applying the common-carrier principle is needed and would probably prevent much of the potential discrimination by network operators. But he is wrong to bundle it with the separations principle. The latter will not stop network operators from discriminating and would almost certainly hamper innovation, while offering only marginal and controversial cultural benefits.
…we need your help. Confronting the many challenges of COVID-19—from the medical to the economic, the social to the political—demands all the moral and deliberative clarity we can muster. In Thinking in a Pandemic, we’ve organized the latest arguments from doctors and epidemiologists, philosophers and economists, legal scholars and historians, activists and citizens, as they think not just through this moment but beyond it. While much remains uncertain, Boston Review’s responsibility to public reason is sure. That’s why you’ll never see a paywall or ads. It also means that we rely on you, our readers, for support. If you like what you read here, pledge your contribution to keep it free for everyone by making a tax-deductible donation.
Vital reading on politics, literature, and more in your inbox. Sign up for our Weekly Newsletter, Monthly Roundup, and event notifications.
How would I know / when I’m empty and quiet like breath?
Historian Gerald Horne has developed a grand theory of U.S. history as a series of devastating backlashes to progress—right down to the present day.
Reflecting on three monumental works of modernism—James Joyce’s Ulysses, T. S. Eliot’s The Waste Land, and Ludwig Wittgenstein’s Tractatus Logico-Philosophicus—a hundred years on.