Equal-Opportunity Internet

Network neutrality also has an unquestioned appeal. The vision of interoperability in which every consumer can access any content, run any application, and attach any device on an equal basis would seem to be a key part of what has made the Internet successful so far.

My core point is that the Internet’s current architecture evolved at a time when the demands that end users were placing on the network were fairly modest and did not vary much. The world today is quite different. End users are placing new demands on the network that call for new solutions.

Three examples illustrate my point nicely.

The first is the emergence of Internet applications that are much more sensitive to delay. With old-style applications, such as e-mail and web browsing, delays of a fraction of a second are virtually unnoticeable. With new-style applications, such as Internet telephony and streaming video, delays of as little as a quarter of a second can destroy a service’s commercial viability. The obvious solution would be to allow networks to give a higher priority to traffic associated with these more time-sensitive applications. Unfortunately, this is precisely the type of discrimination between applications that network neutrality would condemn.

Network neutrality would also have the unfortunate effect of limiting the number of ways in which networks can compete for business. Employing different protocols might permit smaller network players to survive by targeting segments of the market, in much the same way that specialty stores survive in a world increasingly dominated by low-cost, mass-market retailers. In fact, deviating from network neutrality might make it possible for three last-mile networks to coexist: one optimized for traditional Internet applications, such as e-mail and website access; a second incorporating security features to facilitate e-commerce and to guard against problems like viruses and spam; and a third that facilitates time-sensitive applications such as streaming media and VoIP. Adoption of network neutrality would foreclose just such an outcome.

The second example is the appearance of bandwidth-intensive applications, such as music downloading. Published reports estimate that file sharing represents more than half of the world’s Internet traffic. Some broadband providers have considered managing the resulting congestion by placing limits on end-users’ ability to run file-sharing programs. Again, any such restriction would violate the tenets of network neutrality.

The third example is the emergence of content-delivery networks, such as Akamai, which reportedly account for as much as 15% of all of the Internet’s traffic. Suppose that an end user attempts to download a webpage from a leading content provider, such as CNN.com. If CNN.com hosted the content itself, all such queries would be transmitted to the server in CNN’s headquarters in Atlanta. Akamai stores content at a number of locations throughout the Internet. This allows the network to redirect queries to servers that are located more closely and are carrying lighter loads. Doing so reduces delay and opens up new ways to manage the problems of congestion and provides additional protection against denial-of-service attacks. In this respect, content delivery networks represent an important advance that improves the functionality of the Internet. The catch is that Akamai is a commercial service available only to those who are willing to pay for it. And network neutrality would prohibit anything that favors some content over others.

These are just three examples of the ways that new demands on the network are leading Internet providers to experiment with new ways to meet those demands. Together they suggest that Congress and the FCC should consider adopting a policy of promoting network diversity.

What would this mean in terms of Internet policy? At this point, it is impossible to foresee whether network neutrality or network diversity will ultimately represent the better approach. When it is impossible to tell whether a practice would promote or hinder competition, the accepted response to permit the practice to go forward until actual harm to consumers can be proven. Requiring a showing of actual, rather than hypothetical, harm before allowing regulators to intervene provides the room for experimentation upon which the normal competitive process depends.

I look forward to your questions.

Yoo's premise that vigorous competition at every layer is always better for the consumer is overstated. He downplays, to the point of elimination, the basic economic benefits of standardization. . .

Here is an intuitive demonstration of the point. Most people in the United States speak a standard language, English. This undoubtedly leads to some sacrifice. We lose, for example, the precision of German; we lack the Chinese vocabulary for food; and we lose the righteousness and occasional elegance of the French language. But few would argue that vigorous and ongoing competition for a standard American language would clearly serve consumer welfare. It would be, instead, the Tower of Babel.

The same observation holds for standardized technology platforms such as the Windows operating system or the TCP/IP protocol, which bring a wide variety of benefits for application developers and end users. Application writers need only write for a single platform, for example, and can expect to reach a much larger addressable market, thereby justifying greater investments. End-users, given a single standard, share information with ease.

How do you reply to that line of argument?

Christopher Yoo:
    Tim Wu is a friend of mine, and I respect his views immensely and appreciate his willingness to engage my work. I do think that he is misframing the issue slightly.

The undeniable fact that standardization typically yields substantial benefits actually does not justify mandating network neutrality. On the contrary, if the benefits of standardization are sufficiently large, there will be no need to regulate, since all network owners would be expected to adopt it voluntarily.

The real question in deciding whether to make standardization a universal mandate is not whether standardization yields benefits, but rather whether permitting deviations from the existing standard are so dangerous that we should cut them off preemptively without finding out whether such a deviation would be helpful or harmful.

The problem is that deviations from the existing standard can often be quite beneficial. Existing standards change, and Internet policy should provide the means for them to do so.

In addition, universalization of a standard can reduce innovation as well. To amplify on what I said in the Chronicle's main article, Tim often cites the standardization of electric current on 110 volts as an example of the type of standardization he favors. Interestingly, many heavy appliance do not run well at that current level, requiring 220 volt hookups. Most houses include both. This suggests that sometimes what people what from the network and differences in technology's ability to meet those needs may favor some degree of nonstandardization. Indeed, we would lose many beneficial devices if the government were to mandate a uniform electrical standard.

Other examples include cell phone communication protocols. Had the government standardized these early, it would have adopted a standard known as TDMA. A second standard called CDMA has also been deployed that provides different capabilities and provides a better bridge to wireless broadband. Thus, as of today, there are two standards, and yet competition in wireless is generally acknowledged to be quite robust. In fact, a number of leading economic analyses suggest that competition among standards may be more likely to lead to the optimal level of technology adoption and change.

For the first time, network owners would have a strategic capability, as well as incentive, to create barriers to entry for new innovators. We should remember that the current leaders in Internet innovation all began with essentially nothing. Google, eBay, Yahoo! and Amazon all started as simple websites providing limited, but fantastic, services. They had to pay no special access-tax to be on the Internet; there was no special channeling by Internet providers that disadvantage these competitors relative to any others. They succeeded because the product they offered was better than others. Competition on the merits thus drove this market.

That competition would be threatened by access-tiering. Existing content providers have an incentive to block competitors; access-tiering would be a means to effect that competitive advantage. And while these actions might not rise to the level of an antitrust violation, it is perfectly appropriate for Congress to select a network policy that it believes would maximize innovation and growth for the Nation.

How do you reply to that argument?

Christopher Yoo:
    I think that there are risks to innovation on both sides of the network neutrality debate. Many new applications (most notably time-sensitive ones such as Internet telephony, online gaming, and streaming video) depend on the existence of a guaranteed rate of delivery. Without a fast lane, those services will find it much harder to exist. At the same time, it is both logical and fair to ask those applications and content providers whose businesses depend on the network improvements needed to increase the speed of the Internet to help finance those improvements.

Put a different way, I'm sure that we have all had the experience when we were dissatisfied with the quality of the available options and we would have willingly paid more for an option with a better quality of service. My concern is that forcing the Internet into a single level of service would bring that to the Internet.

What is even more interesting is that allowing providers to offer different levels of quality of service can increase the number of ways that network providers can compete. Those of us who recall the early days of wireless telephony remember that cell phone providers the quality of different providers varied widely. Some focused on minimizing dropped calls. Others focused on nationwide coverage. Still others focused on sound quality. And still others focused on price. This type of service differentiation allowed smaller providers to target their network to serve a smaller group of customers who preferred a particular cluster of services, which in turn allowed them to survive in much the same way that specialty stores survive in a world increasingly dominated by low-cost, high-volume discounters. Cutting off competition in terms of service quality would limit network providers to competing on price and network size, which would reinforce the advantages already enjoyed by the largest players. In this way, network neutrality could well have the perverse effect of using regulation to cement the existing last-mile oligopoly into place.

Do you believe that Crawford's scenario is realistic? If so, do you believe that it is a problem?

Christopher Yoo:
    I actually have the opposite concern, i.e., that limiting the Internet to a single class of service will turn the entire Internet into the slow lane. Understanding why is most easily understood by returning to the world before the emergence of the overnight mail services pioneered by Federal Express. Before that, the fastest way to send a letter coast to coast was first-class mail, and that took several days to arrive. FedEx made it possible to send a letter to almost anywhere in the country overnight. The prospect of overnight delivery opened up new ways of doing business that made it possible for businesses to provide a wide range of new services. Providing faster service cost meant higher costs, of course, but those who absolutely, positively needed it to get there overnight were more than happy to pay.

The same arguments apply to the Internet. Allowing for different levels of service promises to enhance innovation by making a wide array of applications that depend on faster delivery. Restricting all traffic to a single class of service would thus have eliminated a great deal of innovation that required faster delivery speeds in order to exist. At the same time, not every letter or data packet needs to travel at the maximum possible speed. Dividing delivery into a premium-priced, faster service and a lower-priced, slower service into a single class allows people who only need the slower speed to pay less for its service. If the delivery service is forced to collapse these two classes of service into a single class, it will probably calibrate the speed of its services to fall somewhere in between and charge a price that falls somewhere in between. Forcing all mail into a single class of service could thus harm consumers by preventing those who don't need the faster service from avoiding having to pay for it.

Sincerely,

Bill Herman

Annenberg School for Communication

University of Pennsylvania

Christopher Yoo:
    This question refers to an article of mine published in the Harvard Journal of Law and Technology. I'm not sure that this is a fair characterization of my work. "Beyond Network Neutrality" only devoted one paragraph to the First Amendment, in which I stated:

"There is nothing incoherent about imposing regulation to promote values other than economic welfare. The problems with this approach are more practical than conceptual."

The point I make in the rest of the paragraph is best understood in light of perhaps the classic First Amendment case on mandating access to a media outlet: the Supreme Court's 1974 decision in Miami Herald v. Tornillo. In that case, the Supreme Court struck down a Florida statute giving candidates a right of reply when the only daily newspaper in town endorsed his opponent.

The Supreme Court basically offered two reasons. First, mandating that the newspaper provide access to speech that it did not control infringed on the newspaper's editorial discretion. Later decisions involving cable and telephone companies make clear that this is the case even when the outlet is simply serving as a conduit for the speech of others.

Second, the Supreme Court recognized that access requirements amount to a tax on speech. Newspapers regard column inches as a precious commodity and allocate them carefully. The concern is that mandatory rights of reply to candidate endorsements means that devoting six column inches to a candidate endorsement really meant having potentially to devote twelve column inches. In effect, devoting space to a candidate endorsement cost the newspaper more than devoting space to other topics. This in turn provides a potential disincentive for a newspaper facing the practical reality of generating enough subscriptions to survive from running candidate endorsements in the first place.

These conclusions suggest that protecting editorial diversity would cut against using network neutrality to give content and applications providers mandatory access to the communication network. At a minimum, it suggests that editorial diversity must be balanced against the economic reality that any media outlet must make enough money to cover its costs if it is to survive.

Club good economics demonstrate the potential problems with the one-size-fits-all pricing that currently dominates the Internet. Consider what would happen if a club charged a flat membership fee that did not vary with club usage. Club members would increase their usage until the costs of exceeded the benefits they would derive from doing so. In the case of flat-fee pricing, the cost to the club members of increasing their usage is zero, so club members will increase their usage until the benefit they derive is zero. The problem is that this increase in usage imposes congestion costs on other club members. This leads to the inefficient outcome that club members will continue to increase their consumption even when the benefits from doing so are more than offset to the cost to the system as a whole. The classic solution is to charge club members a usage-sensitive fee that equals the congestion costs imposed on other users. That provides incentives for club members to set their usage levels at the exact point where the net social benefits equal the net social costs.

This analysis suggests that the evolution toward more complex and usage-sensitive pricing schemes may be nothing more than the natural outgrowth of attempts to manage the increasing congestion on the Internt. It also shows how usage-sensitive pricing can enhance economic welfare.

Needless to say, those concerns never materialized. AOL-Time Warner's business model turned out to be a flop, and all that the merger did was cost the AOL and (particularly) the Time Warner shareholders a lot of money.

The most interesting lesson from this episode are how easy it is to overstate potential harms and how hard it is to foresee which business models will succeed and which ones will fail. It is always possible to hypothesize potential anticompetitive threats. Giving too much weight to potential harms threatens to limit the room for trial and error upon which technological and economic progress depends and may choke off innovation on the basis of potential harms that never materialize.

This is why economic theory and the Supreme Court's antitrust jurisprudence has largely abandoned relying on potential competition as a basis for blocking particular business practices. The conventional wisdom when practices may either be harmful or beneficial is to adopt a wait-and-see approach that allows the business practice to go forward until those challenging the practice can demonstrate actual harm.

The Internet actually works pretty well these days, so it seems dangerous to get Congress monkeying around with its guts. And the question of what constitutes a "neutral" network is pretty subtle, as the term has no widely agreed-upon technical definition.

Having said that, however, s/he proceeded to gripe about the way broadband companies are likely to behave in the absence of network-neutrality rules:

The telecom giants currently sell you Internet access, which is okay, but dull. Dumb pipes are cheap. But hey --- what if they could sell you lots of bundled services? That really gets the dollar signs flashing in their eyes. These are the companies that want to sell ring tone subscriptions for your cell phone, and bundled cable packages with more channels than you'll ever watch. Their dream is to add a dozen extra bullshit services to your Internet service bill, so that you're paying them eighty dollars per month instead of fifty.

The big problem with that plan is that once you have dumb pipes and smart endpoints -- in other words, the Internet -- the endpoints can build essentially any service on top of the network. Of course, this is fantastic for Internet users. Once you pay for Internet access, you automatically get to use every Internet application that's ever been invented: email, the web, instant messaging, peer-to-peer, and (increasingly) the two V's, voice and video. These last two really drive the telecom giants nuts, because they used to sell you voice and video: they're phone and cable companies.

Therefore, the most likely form of telecommunications discrimination in the foreseeable future is discrimination by application, not by content. Verizon wants to give preference to its voice services, and Comcast wants to give preference to its video services. They're deeply freaked out by Skype, Google Video, and the like. If they can convince customers that competing services are slow and crappy compared to their own offerings, they think they'll have a better chance of getting you to pay for their bullshit services. Failing that, they'd like to convince Google and other Internet companies to pay fees for non-degraded service. Think of it as protection money: "Nice customer base you got there. Sure would be a shame if your packets were dropped 20% more often than your competitor's..."

Do you think those fears are well-founded? If so, do you believe that some kind of consumer-protection law (other than network neutrality) might be an appropriate remedy?

Christopher Yoo:
    I would respond in two ways. The first is that the conventional wisdom on bundling has undergone a revolution of sorts over the past fifty years or so, perhaps best explained in terms of the auto industry, which is one of the quintessential vertically integrated industries.

Economic theory and Supreme Court precedent used to be extremely hostile towards bundling. They believed that auto parts manufacturers should be able to sell equally to Ford, GM, and Chrysler. They also believed that local auto dealers should be free to sell cars made by pretty much any manufacturer.

Over time, however, it became clear that allowing for mix-and-match markets at each level in the chain of distribution was only one way to organize an industry. There is also an alternative organizational form, in which dedicated parts manufacturers provide parts exclusively to one of the Big Three, who then sell their cars through dedicated dealerships. Allowing for this type of vertical integration could create numerous efficiencies, such as allowing for custom-designed parts, computerized integration of inventory management systems, and guaranteeing distribution outlets before making multibillion dollar investments in new product lines. Competition under this form is between large vertically integrated enterprises. It is true that you might only be able to buy a particular type of spark plug, transmission, or tire if you bought a particular brand of car. But so long as competition among auto manufacturers is sufficiently robust, it should be sufficient to protect consumers.

The same logic may apply to the Internet. Consider one classic feature of the Internet known as caching. If content is sufficiently popular, a network provider can reduce its costs by storing it locally. This allows it to serve everyone who requests access to that content without having to consumer other network resources, which reduces transmission costs and delay. Of course, caching costs money. The number of requests has to pass a certain minimum threshold before the cost savings from reducing the number of queries justifies incurring the fixed costs to establish and maintain a cache. It thus may make sense for a network provider to guarantee that it reaches this minimum volume by limiting the number of caching services operating at any one point.

Even more importantly, it is widely recognized that bundling cannot possibly harm consumers if there is enough competitive alternatives, since any attempt to force consumers to take both parts of the bundle when they don't want to will only lead them to buy each product separately from competitors. Again, once competition in the last mile becomes sufficiently robust, no network provider will be able to use its market position to use bundling to harm consumers.

When faced with the loss of guaranteed access to the existing network, content and application providers began pouring money into alternative last-mile transmission technologies, such as wireless broadband and broadband over powerline. Most notably, Google has offered to build a wireless broadband network for San Francisco for free.

And I think this is a terrific development. The conventional wisdom in competition policy is that a chain of distribution will only be efficient if every link of the chain is competitive. This suggests that policymakers should focus their efforts on increasing the competitiveness of the link of the chain that is the most concentrated and the most protected by entry barriers (and thus the most likely to remain that way). In the case of the Internet, this link is the last mile.

Google's recent actions show how denying access to the network that exists today creates powerful incentives in investing in alternative last-mile technologies that will constitute the network of tomorrow. Indeed, most network neutrality proponents acknowledge that their concerns will disappear once four last-mile options are available, since any attempt by one of those networks to use exclusivity to harm competition will cause it to lose business to the other three until it returns to procompetitive practices. Unfortunately, the network neutrality debate is framed in terms of its impact on applications and content, which are the portions of the industry that are already the most competitive and the most likely to remain that way. This underscores the extent to which the current debate is focusing on the wrong policy problem.

Your argument here seems to hinge on a user having several options to choose from, in terms of any number of the layers we are concerned about. The fact of the matter, though, is that with the elimination of open access, the last mile is dominated by two groups - telecom and cable. In any number of markets, only one of these may be available, controlled by one company. Independent wireless competition is simply not viable today.

Given the 'pipe' is going to be dominated by one player in many markets, where is this competition in differentiated services going to come from, when the owner of the wire will have ultimate say-so in how its wire is used with the elimination of net neutrality?

This is even to say nothing of effects of consolidation on the backbone of the Internet...

Christopher Yoo:
    I agree that my argument assumes that competition among multiple network providers is feasible. Simply put, the first-order policy goal should be promoting greater competition in the last mile. If that were impossible, it would make sense to pursue the second-order policy goal of promoting complementary services, like applications and content.

That said, the touchstone of the analysis should not be on how many network options we have today, but rather on how many options we will have in the future. The problem is that the policy decisions we make today may affect that in the future.

One of the easiest ways to understand this is terms of the Supreme Court's Terminal Railroad decision, which is the seminal authority for government-mandated access to a bottleneck facility. In Terminal Railroad, there was only one bridge across the Mississippi at St. Louis, and it was owned by a railroad company that was giving preferential treatment to its own rolling stock. The government responded by requiring that the company open up its bridge to all comers. This simply required the bottleneck be shared, and the railroad proceeded to try to charge the monopoly price to everyone. The courts thus had to impose rate regulation, which lowered the price closer to competitive levels.

What is interesting is that the real policy solution was more bridges. The problem is that granting everyone access to the bridge that did exist destroyed any incentive for anyone else to undertake the risk and expense of building another bridge. Even worse, the dampening on investment incentives has the indirect effect of making regulation permanent, since mandating access forestalls the emergence of competition sufficient to justify deregulation.

This is why many commentators (myself included) have stressed the importance of emphasizing dynamic efficiency (i.e., optimizing incentives for creating new resources) over static efficiency (i.e., allocating the resources we have today). The fact that the new bridge may be very expensive and may take a long time is not dispositive, since costly and late may well be better than never.

Would consumers benefit if we forced both sets of travelers into a single class of service? Two things would happen. First, if forced to set one price, airlines would set it somewhere in the middle. This would deprive vacation travelers of the cheaper fares that used to be available. Second, in a nice example of the dynamic efficiency concerns I mentioned above, the drop in total revenue on each flight means that fewer routes will be profitable, and the total number of flights would decrease. Thus attempts to reallocate the flights that do exist may mean that there will be fewer flights to allocate.

I understand that changing from the status quo will create winners and losers. The problem is that right now, consumers of bandwidth intensive applications are imposing costs on low bandwidth users. In fact, the current pricing schemes often effectively require low-bandwidth users to cross subsidize high-bandwidth users. The best way to benefit consumers and to foster the most innovation is to have everyone cover the costs to the network that their activity creates. The high-bandwidth users won't like it much, but it would be better for consumers and for the economy as a whole.

In other words, Jeff, I understand that you are among the most outspoken critics of my work. The debate would do best if we focused on the underlying merits. I hope that we would both welcome disagreement on the issues as normal and healthy, not as a sign that the other side lacks integrity.

Google's recent actions show how denying access to the network that exists today creates powerful incentives in investing in alternative last-mile technologies that will constitute the network of tomorrow. Indeed, most network neutrality proponents acknowledge that their concerns will disappear once four last-mile options are available, since any attempt by one of those networks to use exclusivity to harm competition will cause it to lose business to the other three until it returns to procompetitive practices.

No matter how many last-mile connections exist, eventually content traveling over these connections will need to cross the country via middle-mile and backbone connections owned by rapidly-consolidating telco and cable giants. We currently pay little regulatory attention to these pipes.

What good is last-mile competition if content from San Francisco to Philadelphia might be slowed down by AT&T in the middle if it so decides, given open access don't apply and net neutrality rules may not either soon?

Christopher Yoo:
    I meant to mention the middle mile and backbone providers in my answer to your earlier question.

I guess I don't share your concren. By current standards applied by the Justice Department, FTC, and FCC, the backbone market is actually fairly competitive. The exact number of backbone providers isn't completely clear, but there seem to be somewhere around five. That should be enough to discipline anyone who attempted to harm competition. Any frustrated content or applications provider would find four other backbones ready to welcome it with open arms.

In addition, there are few entry barriers preventing a new company from laying a long-haul fiber network. So, again, I would direct our attention to the link that is more concentrated and protected by entry barriers: the last mile. Once you get into the details, there are some smaller issues about backbones that may need addressing, but those require fixes that are much more tailored and less broad-brush than network neutrality.

I'm sure that we have all had the experience when we were dissatisfied with the quality of the available options and we would have willingly paid more for an option with a better quality of service. My concern is that forcing the Internet into a single level of service would bring that to the Internet.

But Larry Lessig denies that network-neutrality would have that effect. In his Senate testimony, he said that he thinks "consumer tiering" -- in which broadband companies charge their end-users higher rates for higher connection speeds -- is perfectly fine:

To oppose access-tiering, however, is not to oppose all tiering. I believe, for example, that consumer-tiering should be encouraged. Network providers need incentives to build better broadband services. Consumer-tiering would provide those incentives.

In other words, Lessig believes that it's fine to charge consumers different rates for different speeds, but it's a problem to charge application and content providers different rates for different speeds.

Is Lessig wrong about the feasibility of this? Why do you believe that network neutrality would inevitably lead to a single tier of service at the consumer end?

Christopher Yoo:
    The Internet is a classic example of what is called a two-sided market, in that it serves as an intermediary that brings content and applications providers and subscribers by entering into separate contracts with both sides.

I was glad to see that Prof. Lessig has embraced what he calls consumer tiering. My problem is that what seems good for the goose also seems good for the gander. If differential pricing for heavy-bandwidth users makes sense on the consumer side, why doesn't it also make sense on the content and applications provider side?

Put a different way, network owners who are enhancing the current Internet must be able to generate additional revenue either from consumers or from applications and content providers, otherwise they won't invest in the enhancements. If network providers are able to engage in differential pricing on both the access and consumer sides, they will calibrate pricing in both directions to try to reach the best possible solution. If they cannot engage in differential pricing on the access side, however, they will be prevented from using pricing to place the costs of the network enhancements on the content and applications providers who benefit from them, and they will have no choice but to direct more of that burden onto consumers. I see no reason to give the content and applications providers such a free ride.

So yes, the Alaska folks are in a quandary, as are we all. But it is a quandary not much different (and perhaps more justifiable) than the one faced by every city that had to run school buses after the recent run-up in gasoline prices. What I am not convinced is that the solution is locking the Internet into a one-size-fits-all architecture. Interestingly, the Internet is currently designed are a thirty-year-old suite of protocols that many technologists regard as obsolete. If so, it would be a big mistake to freeze it into place.

In many ways, I regard most of the fight as being an intramural fight between large Internet players that has fewer implications for consumers than most initially think. The amount that cable modem and DSL providers can charge you and me is determined largely by the lack of last-mile alternatives. Even if the government were to mandate network neutrality, I wouldn't expect those prices to change much. I would still effectively only have two broadband options. Network neutrality would have a dramatic effect on how the Googles and the Verizons of the world divide up those profits, which would normally be divided in accordance with their relative bargaining power. Both sides are instead trying to use the regulatory process to gain the upper hand in that fight. While of intense interest to those companies and their shareholders, the division of profits among upstream players does not affect consumers. I thus question whether network neutrality deserves as much focus among policymakers as it has received.