Electronic Comment Filing System

December 14, 2010 ELECTRONIC FILING Chairman Julius Genachowski Commissioner Michael J. Copps Commissioner Robert M. McDowell Commissioner Mignon Clyburn Commissioner Meredith Attwell Baker Federal Communications Commission 445 12th Street S.W. Washington, DC 20554 Re: Preserving the Open Internet, GN Docket No. 09-191; Broadband Industry Practices, WC Docket No. 07-52; Framework for Broadband Internet Service, GN Docket No. 10-127 Dear Chairman Genachowski and Commissioners: Attached please find attached my white paper entitled ?Network Neutrality: What a Non- Discrimination Rule Should Look Like.? The paper discusses alternative proposals for non- discrimination rules and outlines my proposal for a non-discrimination rule that would ban all application-specific discrimination (i.e. discrimination based on application or class of application), but allow all application-agnostic discrimination. Sincerely, /s/ Barbara van Schewick Barbara van Schewick Associate Professor of Law and (by courtesy) Electrical Engineering Faculty Director, Center for Internet and Society Stanford Law School 650-723-8340 schewick@stanford.edu Barbara van Schewick Network Neutrality - What a Non-Discrimination Rule Should Look Like Version 1.0 - December 14, 2010 Center for Internet and Society White Paper Network Neutrality: What A Non-Discrimination Rule Should Look Like Barbara van Schewick1 Center for Internet and Society White Paper Version 1.0, December 14, 2010 Executive Summary The debate over "network neutrality," i.e. the debate over whether governments should establish rules limiting the extent to which network providers can interfere with the applications and content on their networks, has become one of the hottest debates in Internet policy. Governments all over the world, including the European Union, the UK, France, Germany and the US, are investigating whether regulatory action is needed. Beyond rules that prevent network providers from blocking applications or content, non-discrimination rules are a key component of any network neutrality regime. There is, however, a lot of uncertainty about how to best distinguish harmful from beneficial discrimination. The paper sets out criteria for evaluating non-discrimination rules and uses these criteria to evaluate the following proposals for a non-discrimination rule: ? The decision not to have a non-discrimination rule. This would allow all forms of discrimination and impose no constraints on a network provider?s ability to offer Quality of Service. ? A non-discrimination rule that would ban all discrimination. This rule would prohibit all forms of Quality of Service. ? A non-discrimination rule that would ban discrimination based on application, but would allow discrimination based on class of application and application-agnostic discrimination. This rule would allow forms of Quality of Service that respect ?like treatment.? ? A non-discrimination rule that would ban all application-specific discrimination (i.e. discrimination based on applications or classes of applications), but would allow application-agnostic discrimination. This rule would allow certain forms of user- controlled Quality of Service, but not all forms of Quality of Service. 1 Associate Professor of Law and (by Courtesy) Electrical Engineering, Stanford Law School; Director, Center for Internet and Society, Stanford Law School. An earlier version of this paper was filed with the Federal Communications Commission in August 2010 as an attachment to a notice of an ex parte conversation. That version is available at http://fjallfoss.fcc.gov/ecfs/document/view?id=7020652515. A subsequent version was presented at the 38th Research Conference on Communication, Information and Internet Policy (TPRC 2010). That version is available at: http://www.tprcweb.com/images/stories/2010%20papers/van%20Schewick%20TPRC%202010%20What%20a%20 non-discrimination%20rule%20should%20look%20like.pdf . van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 2 ? A non-discrimination rule that would ban discrimination that causes harm to users or harm to competition. Whether these conditions (anticompetitive, harm to users) are met would be decided by the regulatory agency in case-by-case adjudication. Such a rule was part of the Google-Verizon legislative proposal. Regulators or legislators should adopt a non-discrimination rule that clearly bans application-specific discrimination (i.e. discrimination based on application or class of application), but allows application-agnostic discrimination.2 Such a rule protects the factors that have fostered application innovation in the past, ensuring that the Internet can continue to serve as an engine of innovation and economic growth in the future. It preserves the factors that have allowed the Internet to improve democratic discourse and to provide a decentralized environment for social and cultural interaction in which anyone can participate. This rule would prevent network providers from distorting the playing field between applications or classes of applications. It would provide certainty to all market participants. Network providers would know how they can manage their networks, and application developers (and their investors) could be sure that they won?t be discriminated against. The rule does not constrain the evolution of the network more than is necessary to reach the goals of network neutrality regulation. In particular, it allows certain forms of user-controlled Quality of Service. Finally, it gives network providers the tools they need to manage their networks, while preserving application innovation and user choice as much as possible. The rule would be coupled with an exception for reasonable network management that would require reasonable network management to be as application-agnostic as possible. 2 Throughout this paper, I use ?applications? as a shorthand for ?application, content and services.? van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 3 INTRODUCTION The debate over "network neutrality," i.e. the debate over whether governments should establish rules limiting the extent to which network providers can interfere with the applications and content on their networks, has become one of the hottest debates in Internet policy. Governments all over the world, including the European Union, the UK, France, Germany and the US, are investigating whether regulatory action is needed. Regulators considering whether to enact network neutrality rules need to answer a series of questions in order to decide which, if any, network neutrality rules they should adopt. The first question is: ?Do we need a rule against blocking, i.e. a rule that prohibits network providers to block applications and content on their networks?? Such a rule is part of all network neutrality proposals; this is the one rule on which all network neutrality proponents agree. This paper assumes that the case for a rule against blocking has been made. But what about behavior that stops short of blocking? For example, if Comcast slows down Internet video applications like Hulu or YouTube that compete with Fancast, Comcast?s own Internet video application, should this be prohibited, too? In the context of the network neutrality debate, the term for this type of behavior that stops short of blocking is ?discrimination,? and the question regulators need to answer is: ?Should we only ban blocking, or also discrimination, and if we ban discrimination, how should discrimination be defined?? Answering this question is the goal of this paper. It sets out criteria for evaluating non-discrimination rules and uses these criteria to evaluate the following proposals for a non-discrimination rule: ? All or nothing approaches: o Allow all discrimination. o Ban all discrimination. ? Approaches protecting the application-blindness of the network: o Ban all discrimination based on application, but allow discrimination based on class of applications and application-agnostic discrimination. o Ban all discrimination based on application or class of applications, but allow all application-agnostic discrimination. ? Other approaches: o Ban discrimination that is anticompetitive or harms users. o Ban discrimination that is not disclosed. The paper concludes that network neutrality rules should ban blocking AND discrimination, not just blocking. They should ban all discrimination based on application or class of application, but allow all forms of application-agnostic discrimination. This rule would van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 4 allow network providers to offer different classes of services, if they meet the following conditions: (1) the different classes of service are offered on a non-discriminatory basis, i.e. without regard to the identity of the sender or receiver, the specific application or content (e.g. Skype vs. Vonage), or the class of application or content (e.g. Internet telephony vs. e-mail); (2) the user is able to choose whether and when to use which class of service; and (3) the network provider is allowed to charge only its own Internet service customers (not application and content providers who are not its Internet service customers) for the use of the different classes of service. CRITERIA FOR EVALUATING NON-DISCRIMINATION RULES A non-discrimination rule should meet the following criteria: ? It should protect the factors that have fostered application innovation in the past to ensure that the Internet can continue to serve as an engine of innovation and economic growth in the future. They are:3 ? Innovator choice: Innovators independently choose which applications they want to pursue; they do not need support or ?permission? from network providers in order to realize their ideas for an application (this factor has also been called ?innovation without permission?). Adding additional decision makers who need to endorse the idea or take action before an idea can be realized reduces the chances that innovative ideas can be realized. 4 ? User choice: Users independently choose which applications they want to use, without interference from network providers. Letting users, not network providers choose which applications will be successful is an important part of the mechanism that produces innovation under uncertainty. 5 At the same time, letting users choose how they want to use the network enables them to use the Internet in a way that creates more value for them (and for society) than if network providers made this choice.6 3 These factors are described in detail in van Schewick (2010a). For a short overview, see van Schewick (2010c). 4 On innovation without permission in the original Internet, see , pp. 204, 211, 293. On the impact of innovation without permission on innovation, see van Schewick (2010a), pp. 345-348. 5 See van Schewick (2010a), pp. 349-351 and van Schewick (2010c), p. 6. 6 See van Schewick (2010a). pp. 362-363: From van Schewick (2008), pp. 7-8: ?Why the emphasis on user choice? First, user choice is fundamental if the Internet is to create the maximum value to society. The Internet is a general purpose technology. It does not create value through its existence alone. It creates value by enabling users to do the things they want or need to do. Users know best what this is. As a result, users, not network providers should be able to decide how they would like to use the network, and what is important to them. Of course, in order for users to van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 5 ? Application-Blindness: The application-blindness of the network ensures that network providers cannot interfere with these (i.e. innovators? and users?) choices, that they cannot distort competition among applications (or classes of applications) or reduce application developers? profits through access charges (we may call this ?innovation without fear?).7 ? Low costs of innovation: The low costs of innovation not only make many more applications worth pursuing, 8 but also allow a large and diverse group of people to become innovators, which in turn increases the overall amount and quality of innovation.9 ? It should protect the factors that have allowed the Internet to improve democratic discourse and to provide a decentralized environment for social and cultural interaction in which anyone can participate. 10 ? It should not constrain the evolution of the network more than is necessary to reach the goals of network neutrality regulation. behave efficiently, they also need to bear (at least some of) the costs of their actions, something which the current system does not sufficiently provide. User choice is also a fundamental component of the mechanism that enables application-level innovation to function effectively. In the current Internet, it is impossible to predict what future successful applications will be. Enabling widespread experimentation at the application-level and enabling users to choose the applications they prefer is at the heart of the mechanism that enables innovation under uncertainty to be successful. By singling out specific applications, network providers start picking winners and losers on the Internet. As we have seen, whom they pick may be driven by a number of motivations that are not necessarily identical with what users would prefer, leading to applications that users would not have chosen and forcing users to engage in an Internet usage that does not create the value it could. Consumers, not network providers, should continue to choose winners and losers on the Internet.? 7 A network is application-blind, if it cannot distinguish between the applications and content on the network. The original Internet was application-blind, which was a consequence of applying the broad version of the end-to-end arguments. See van Schewick (2010a), pp. 72-75, 217-218. For a short summary of the importance of application- blindness, see van Schewick (2010c), pp. 3-4. For a detailed analysis of the impact of application-blindness on network providers? incentives to discriminate and on the pricing strategies available to network providers, see van Schewick (2010a), chapter 6. 8 For a short version of the argument, see van Schewick (2010c), pp. 2-3, 5-6 and van Schewick (2010b), pp. 4-5. On low cost of application innovation in the original Internet, see van Schewick (2010a), pp. 138-148. On the impact of low cost innovation on who can innovate, see van Schewick (2010a), pp. 204-213. 9 If there is uncertainty (e.g., about technology or user needs) or user needs are heterogeneous, a larger and more diverse group of innovators will create more and better application innovation than a smaller, less diverse group of innovators, and these applications will better meet the needs of Internet users. For the short version of the argument, see van Schewick (2010c), pp. 5-6 and van Schewick (2010b), pp. 4-5. For the detailed version, van Schewick (2010a), pp. 298-349. In the current Internet, there is uncertainty and user needs are heterogeneous, so the conditions under which innovator diversity increases the amount and quality of innovation are met. See van Schewick (2010a), p. 356. 10 For a brief discussion of these factors, see van Schewick (2010a), pp. 359-365. van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 6 ? It should make it easy to determine which behavior is and is not allowed to provide much-needed certainty for industry participants. ? It should keep the costs of regulation low. EVALUATING ALTERNATIVE PROPOSALS FOR NON-DISCRIMINATION RULES When determining whether to adopt network neutrality rules, legislators and regulators need to decide whether the network neutrality rules should only ban blocking, or also discrimination, and if they decide to ban discrimination, how discrimination should be defined. The answers to these questions may affect how the core of the network can evolve. In particular, they determine whether a network provider can offer Quality of Service (see box 1). Box 1: Quality of Service In the original Internet, the network provides a single best-effort service. That is, the network does its best to deliver data packets, but does not provide any guarantees with respect to delay, bandwidth or losses. Thus, the network operates like the default service offered by the postal service, which does not guarantee when your letter will arrive or whether it will arrive at all. Contrary to the postal service, which lets users choose services other than the default service like two-day shipping, the original Internet provides only best-effort service. But different applications have different needs. For example, Internet telephony is very sensitive to delay, but does not care about occasional packet loss. By contrast, e-mail is very sensitive to packet loss, but does not care about some delay. So you could imagine a network that treats packets belonging to different applications differently, depending on their needs. For example, a network could give low delay service to Internet telephony packets, but best-efforts service to e- mail packets. A network that offers different types of service to different data packets is a network that offers ?Quality of Service.? All- or Nothing-Approaches A first set of approaches takes an all-or-nothing position towards differential treatment of packets. Option: Allow All Discrimination (or ?No Rule Against Discrimination?) Network providers and other opponents of network neutrality regulation oppose any restrictions on network providers? ability to differentiate among data packets. According to them, a ban on discrimination would make it impossible to offer supposedly useful services such as Quality of Service, or to manage networks during times of congestion. Without Quality of Service, proponents of this option argue, certain types of applications (those that require special treatment van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 7 from the network) will not be able to operate. Thus, banning Quality of Service may reduce application innovation. While some forms of differential treatments such as those involved in Quality of Service would be socially beneficial, the argument continues, trying to distinguish between beneficial and harmful discrimination (to the extent it exists) would be too difficult. Since technology is evolving rapidly, regulators are likely to get it wrong. Even if regulators succeed in identifying criteria that accurately distinguish between beneficial and harmful discrimination when the regulation is enacted, these criteria may not be accurate in the future. This view neglects that banning blocking, but allowing discrimination will make the rule against blocking meaningless by offering a legal alternative to blocking that is less costly and potentially more effective. Discrimination is an attractive alternative to blocking. Discrimination reduces the perceived quality of the affected application relative to others. If a network provider secretly slows down packets or uses methods that are difficult to detect, its customers may attribute the affected application?s or website?s bad performance to bad design, and happily switch to the network provider?s supposedly superior offering. While the result of blocking and discrimination is the same - the network provider?s Internet service customers stop using the blocked or degraded application and switch to the affiliated application -, the costs of using discrimination instead of blocking are much lower. If the network provider blocks an application, users will notice and may switch to another Internet service provider. By contrast, users who do not realize that their network provider interfered with their preferred application and think they chose the better application will have no incentive to switch.11 Based on these considerations, we would expect network providers to prefer discrimination over outright blocking. Network providers? actual behavior is in line with these predictions. In the examples that are often highlighted in the debate, network providers often used methods that make it more difficult or costly to reach particular applications or content instead of blocking access to them completely. For example, Bell Canada only throttles, but does not block peer-to-peer file sharing applications. In 2009, BT only restricted the bandwidth available to the BBC iPlayer or YouTube videos (in BT?s Option 1 Broadband service). Available evidence suggests that network providers are well aware of the advantages of this strategy. To shut down BitTorrent connections, Comcast used ?forged? data packets that seemed to come from the other party involved in the specific BitTorrent connection. As white papers produced by Sandvine, the equipment vendor Comcast was using, showed, this method of interference was deliberately chosen to prevent customers from noticing the interference. 12 11 For a variety of reasons, requiring network providers to disclose whether they interfere with applications and content will not fully solve this problem. See the brief discussion at the end of this paper. Network providers know that the use of file-sharing applications is an important driver of 12 Sandvine (2004), p. 14. van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 8 broadband adoption, and they do not want to lose customers who wish to use these applications.13 Thus, an effective network neutrality regime needs to ban blocking and discrimination. Beneficial forms of blocking or discrimination can be accommodated through the definition of discrimination or through exceptions. Ban all discrimination By contrast, some participants in the debate would ban all discrimination, requiring network providers to treat every packet the same.14 Proponents of this option are concerned that network providers may use the provision of Quality of Service to distort competition among applications or classes of applications. They point out that network providers who offer Quality of Service have an incentive to reduce the quality of the baseline service below acceptable levels, and that offering Quality of Service may enable network providers to profit from bandwidth scarcity instead of motivating them to increase the capacity of their networks. While these arguments all have merit, these problems can be solved without totally banning Quality of Service. As will be explained below, it is sufficient to constrain how it can be offered. The FCC?s draft rules in the Open Internet Proceeding are an example of this type of non-discrimination rule. A rule that required network providers to treat every packet the same would make it impossible to offer Quality of Service, which, by definition, entails the network treating packets differently. Proponents of this option also question the need for Quality of Service. They argue that so far, the lack of Quality of Service in the public Internet has not prevented real-time applications from becoming successful. For example, although Internet telephony is sensitive to delay and may benefit from a network service that guarantees low delay, Internet telephony applications such as Skype or Vonage work well in the current Internet. Cisco?s real-time video- conferencing service for home users called ??mi telepresence? is now offered to home users. According to Cisco, the service works with broadband Internet connections that provide at least 1.5 mbps to and from the home; it does not require special treatment from the network.15 13 Mennecke (2005); Hellweg (2003); Sandvine (2004), pp. 5-6. See also The National Cable & Telecommunications Association (2007), p. 31 (?cable operators will not go down the path of blocking access to video or P2P services. Blocking such services would be recipe for [?] massive dissatisfaction among consumers, which would lead to loss of customers to our competitors.?) The success of real-time applications despite the lack of Quality of Service seems to be due to two factors: in many regions the current Internet seems to have sufficient capacity to prevent the lack of Quality of Service from becoming a problem. In addition, network engineers and application designers have developed techniques that help real-time applications cope with the problems 1414Internet Non-Discrimination Act of 2006 (2006); Crawford (2007), pp. 403-404. 15 http://home.cisco.com/en-us/telepresence/umi/what-you-need (visited November 26, 2010); http://home.cisco.com/en-us/telepresence/umi/what-you-need (visited November 26, 2010) van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 9 resulting from best-effort service. While it is unclear whether the lack of Quality of Service in the public Internet is currently preventing certain types of real-time applications from developing, it is possible, as a theoretical matter, that the lack of Quality of Service may preclude certain types of applications that have stricter requirements than the current best-effort Internet can provide. Thus, it is at least possible that a total ban on Quality of Service may reduce application innovation. Finally, proponents of a ban point out that in the networking community, the value of Quality of Service is not undisputed. Providing Quality of Service makes the network more complex. Network engineers debate whether the added complexity is worth the effort, or whether increasing network capacity (a solution that is called ?overprovisioning?) would be preferable. If there is enough capacity, there is no congestion and the level of delay is low enough to be tolerable for real-time applications. For example, after trying for several years to implement Quality of Service mechanisms in the research network Internet2, Internet2 researchers suspended the effort indefinitely and concluded that adding more capacity was the better solution. In congressional testimony, representatives of Internet2 have pointed to this experience to negate the need for Quality of Service.16 In sum, while there are legitimate concerns about the consequences of allowing Quality of Service on competition among applications or investment in the network, these concerns can be mitigated without totally banning Quality of Service. At the same time, it is at least possible that a total ban on Quality of Service would harm applications that need strict guarantees from the network. Under these circumstances, requiring network providers to treat every packet the same would be too restrictive, constraining the evolution of the network more than absolutely necessary to protect the values that network neutrality is designed to protect. Whatever the merits of this debate from an engineering perspective, these arguments should be irrelevant for the regulatory debate over network neutrality rules. Network neutrality rules are the result of a trade-off. They impose some constraints on the evolution of the network in order to allow the Internet to continue to foster application innovation, preserve user choice or foster democratic discourse. Legislators and regulators need to decide whether restrictions on the evolution of the network are necessary to protect the values that network neutrality rules are designed to protect. By contrast, the question of whether certain changes to the network make sense from a technical or business perspective is a question that should be left to network engineers and network providers. Approaches Protecting the Application-Blindness of the Network A second set of approaches seeks to preserve the application-blindness of the network which the Internet exhibited in the past. Contrary to the approaches discussed above which ban or allow all discrimination, the approaches in this group ban only some, but not all discrimination. 16 http://qos.internet2.edu/wg/documents-informational/20020503-premium-problems-non-architectural.html; http://net.educause.edu/ir/library/pdf/EPO0651.pdf. van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 10 An application-blind network is unable to distinguish among the applications and content on a network, and, as a result, is unable to make distinctions among data packets based on this information. The original Internet was application-blind, which was a consequence of its architecture. As I have explained in detail elsewhere, the Internet?s application-blindness is one of the factors that have fostered application innovation in the past and made the Internet more valuable for users and for society. It also contributed to the Internet?s ability to improve democratic discourse, facilitate political organization and action, and create a decentralized environment for cultural and political interaction in which anybody can participate. Today, technologies such as Deep Packet Inspection have removed the application-blindness of the network. They allow network providers to identify the applications and content on their networks and to control their execution. Thus, the non-discrimination rules in this category try to preserve the application- blindness of the network in order to preserve the Internet?s ability to function as a general- purpose platform over which applications, content and services compete on a level playing field, with users choosing which applications become successful and how the network can be used. The first, less restrictive approach, only bans discrimination based on the specific application or content or the identity of the application or content provider (i.e. treating Vonage differently from Skype, or treating the website of the New York Times differently from the website of a community group), but allows discrimination based on the class of application (i.e. treating Internet telephony differently from e-mail) and discrimination based on application- agnostic criteria. The second approach, which is the one proposed by this paper, bans discrimination based on application and class of application, but allows all application-agnostic discrimination. Ban all discrimination based on application, but allow discrimination based on class of application and application-agnostic discrimination The first approach only prohibits discrimination based on the specific application, but allows discrimination based on the class of application and all application-agnostic discrimination. [Again, I use ?applications? as a shorthand for ?application, content and services?.]Thus, under this approach, a network provider would not be allowed to treat Hulu different from YouTube, or Fancast differently from Hulu.17 This part of the rule is designed to prevent network providers from treating specific applications differently as a substitute for blocking. In all cases in which a network provider may have an incentive to block one or more applications to favor another application, it would have an incentive to reach the same result by treating the favored application relatively better than the others (either by treating the favored application better or the other applications worse), and the rule is designed to prevent this. Providing Quality of Service only to one application, but not others would be one example of this behavior. This would be discrimination based on application. 17 Thus, it does not matter whether the favored application is affiliated with the network provider or not. van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 11 This approach would, however, allow network providers to discriminate based on the class of applications as long as they treat like traffic alike (this is often called ?like treatment?). Under rules that require ?like-treatment,? network providers are required to treat like traffic alike: they are allowed to treat classes of applications differently, as long as they do not discriminate among applications within a class.18 Since network providers have to treat applications with different needs in the same way, ?like treatment? would, proponents of this approach argue, make it impossible for network providers to use the provision of Quality of Service to distort competition among applications with similar needs. At the same time, they see no harm in allowing network providers to treat applications with different needs differently. For example, e-mail and Internet telephony have different requirements with respect to reliability and delay: e-mail requires reliable data transfer, but is not sensitive to delay. By contrast, Internet telephony can deal with a certain amount of packet loss, but is very sensitive to delay. As a result, it does not harm e-mail if the network provider gives low delay service to Internet telephony, but not to e-mail. For example, under rules that require like- treatment, network providers would be allowed to treat Vonage, an Internet telephony application, different from Gmail, an e-mail application, but they would not be allowed to treat Skype, another Internet telephony application, differently from Vonage. This view neglects that like treatment negatively affects several of the factors that have fostered application innovation in the past. First, allowing network providers to treat classes of applications differently requires the network provider to identify the different applications on its network in order to decide which class they belong to and determine the appropriate form of Quality of service, removing the application-blindness of the network. The concept of ?like applications? is not well defined. This fuzziness enables network providers to distort competition among applications or classes of applications. For example, a network provider may fail to provide the needed type of service to a certain application in a class, arguing it had not realized that this application belonged to this class. Alternatively, a network provider may argue that an application does not belong to a certain class. The Canadian Radio-television and Telecommunications Commission?s (CRTC) review of the Internet traffic management practices of Internet service providers19 18 On like treatment, see Wu (2006b), pp. 42-43. The merger conditions in the AT&T/BellSouth merger allowed this type of QoS. See Wu (2006a). provided an example of this type of problem. The proceeding showed that many Canadian Internet service providers were slowing down traffic belonging to peer-to-peer file-sharing applications all day or during times of congestion. They argued that this was necessary in order to protect the performance of real-time applications (such as applications that stream video in real time) during times of congestion. This raised an interesting question: How do the network providers treat Vuze, an application that, at the time of the proceeding, used the BitTorrent protocol, a peer-to-peer file-sharing protocol to stream video in real time? Is it 19 Notice of Consultation and Hearing available at http://www.crtc.gc.ca/eng/archive/2008/pt2008-19.htm; Record of the proceeding available at http://www.crtc.gc.ca/PartVII/eng/2008/8646/c12_200815400.htm. van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 12 treated like the other peer-to-peer file-sharing applications, or like other applications that stream video in real time? The first option would put Vuze at a competitive disadvantage compared to applications like YouTube that also stream video on real time, but are based on a client-server architecture. The record of the proceeding did not answer this question, and, although several groups brought a complaint to get the answer to this question, the CRTC declined to explore the issue further.20 In addition, a network provider could define classes of applications in a way that distorts competition among classes of applications. For example, network providers usually like the idea of providing low delay service to online gaming. Some online games are sensitive to delay, and this would allow network providers to capture some of the value that online gamers realize from gaming by charging for the special type of service. By contrast, network providers seem to be less interested in providing low delay service to Internet telephony applications like Skype or Vonage, since this would make these applications more competitive with the network providers? own telephony offerings. Thus, a network provider may decide to offer low delay service only to online gaming, but not to Internet telephony, arguing that these are different classes applications. Internet telephony providers would argue that the correct class would be ?applications that are sensitive to delay,? and I think that is right, but it would not matter until they had brought a complaint and succeeded in convincing the regulatory agency. As these examples show, disputes over which classes of applications are alike, or whether a certain application belongs to a certain class, are likely to be frequent and difficult to resolve, creating high costs of regulation. Under ?like treatment?, network providers, not users choose which application should get which Quality of Service (violating the principle of user choice). If the network provider decides whether and when to offer QoS, it is forced to guess what the average user's priorities may look like, but these priorities may differ among users, and (for the same user) over time. In particular, a specific user?s needs with respect to a particular application are not necessarily fixed:21 Finally, ?like treatment? harms application innovation by making it more difficult for new applications to get the type of service they need. In order to get QoS, an application developer would have to convince network providers that its new application is a new class of A user?s desire for QoS may differ considerably depending on the circumstances. For example, I may not care as much about the quality of my VoIP call when I'm chatting with a friend as when I'm doing a job interview. If I?m playing a quick game at night I may be willing to tolerate a level of latency that I would not be willing to tolerate during an online gaming tournament. Thus, any QoS system that lets network providers determine whether and when to provide QoS may not be aligned well with users? needs. 20 Application for review at http://www.crtc.gc.ca/PartVII/eng/2009/8662/p8_200907727.htm; CRTC decision at http://www.crtc.gc.ca/eng/archive/2009/2009-677.htm (relevant discussion under para 8-10). 21 See also Briscoe, Moncaster & Burness (2007), section 3.3. van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 13 application that requires QoS or that it is ?like? an existing type of application that already receives QoS. This introduces considerable transaction costs. Certain types of innovators (e.g. non-commercial innovators, innovators that develop an application at home in their free time or start ups) may not have the resources necessary to engage in this type of negotiation with a potentially large number of network providers. In addition, even if an innovator manages to contact a network provider, the innovator may not receive the appropriate QoS for its application if the innovator fails to convince the network provider. This is an example of the more general phenomenon ? described at the workshop in January ? that requiring cooperation or support from the network provider reduces the likelihood that innovative applications can be realized or successfully deployed.22 Thus, requiring network providers to take action before an application can get the Quality of Service it needs violates the principle of innovation without permission and reduces the chance that new applications actually get the type of service they need. Ban all discrimination based on application or class of application, but allow all application-agnostic discrimination Instead, regulators or legislators should adopt a non-discrimination standard that clearly bans application-specific discrimination (i.e. discrimination based on application or class of application), [again, I use ?applications? as a shorthand for ?application, content and services?] but allows application-agnostic discrimination. Thus, a network provider would not be allowed to treat Vonage differently from Skype, or Comcast?s Fancast differently from Hulu. That would be discrimination based on application. Nor would it be allowed to treat online video differently from e-mail. That would be discrimination based on class of application. But it would be allowed to treat data packets differently based on criteria that have nothing to do with the application or class of application. For example, during times of congestion, a network provider could give one person a larger share of the available bandwidth than another, for example because this person pays more for Internet access or has used the Internet less over a certain period of time. But it could not throttle the bandwidth available to a specific online video application such as Hulu in particular, or online video in general. That would be application-specific discrimination. Protecting the factors that are at the core of the Internet?s economic, social, cultural and political potential Such a rule preserves the application-blindness of the network, the principle of user choice, and the principle of innovation without permission, three factors that have been central to the Internet?s ability to foster innovation in the past.23 22 See the references under ?innovation without permission? in the section ?CRITERIA FOR EVALUATING NON-DISCRIMINATION RULES? above. By prohibiting application-specific 23 See van Schewick (2010a). For a short summary, see van Schewick (2010c). van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 14 discrimination, the proposed rule makes it impossible for network providers to distort competition among applications or classes of applications. The rule provides certainty all market participants. Network providers know how they can manage their networks. Application developers and their investors know that they will have a fair chance in the market place ? that they will be able to reach users and compete with other applications on the merits, without interference from network providers. The rule allows users, not network providers to choose how they want to use the network and which applications will be successful. Letting users make this choice not only increases the value of the Internet for users and for society, it is also an important part of the mechanism that enables application-level innovation to function effectively. In addition, maintaining application-blindness and user choice is key to allowing the Internet to realize its social, cultural and political potential. Allowing the network to evolve The proposed rule does not constrain the evolution of the network more than is necessary to reach the goals of network neutrality regulation. It provides room for networks to evolve. Quality of Service The rule allows network providers to offer certain (though not all) forms of Quality of Service. In particular, it allows network providers to offer different classes of service if they meet the following conditions: (1) the different classes of service are offered on a non-discriminatory basis, i.e. without regard to the specific application or content or the specific application or content provider (e.g. Skype or Vonage), or the type of application or content (e.g. Internet telephony); (2) the user is able to choose whether and when to use which class of service; (3) the network provider is allowed to charge only its own Internet service customers for the use of the different classes of service.24 This type of user-controlled Quality of Service offers the same potential societal benefits as other, discriminatory or provider-controlled forms of Quality of Service without the social costs. In particular, it does not raise any of the problems associated with ?like treatment.? Unlike like treatment, it preserves the application-blindness of the network, the principle of user choice, and the principle of innovation without permission: First, the proposal maintains the application-blindness of the network: The provision of Quality of Service is not dependent on which applications users are using, but on the Quality-of- Service-related choices that users make; thus, the network providers does not need to know 24 This rule would not constrain interconnection agreements in any way. I explained the rationale for this criterion in van Schewick (2010b). While the first two conditions are a consequence of the proposed non-discrimination rule, the third condition would have to be encoded separately. van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 15 anything about which applications are using its network in order for this scheme to work. The network provider only makes different classes of service available, but does not have any role in deciding which application gets which Quality of Service; this choice is for users to make. As a result, network providers cannot use the provision of Quality of Service as a mechanism to distort competition among applications or classes of applications. Second, since users choose when and for which applications to use which type of service (in line with the principle of user choice), they can get exactly the Quality of Service that meets their preferences, even if these preferences differ across users or (for a single user) over time. Third, in line with the principle of ?innovation without permission,? an innovator does not need support from the network provider in order for his application to get the Quality of Service it needs. The only actors who need to be convinced that the application needs Quality of Service are the innovator, who needs to communicate this to the user, and the user, who wants to use the application. This greatly increases the chance that an application can get the type of service it needs. Finally, under the proposed rule, the costs of regulation are lower than under a rule that would allow QoS as long as it provides like treatment, because the proposed rule does not suffer from definitional ambiguities and does not offer similar possibilities for gaming the system. Network Management The proposed rule allows network providers to freely engage in application-agnostic ways of managing congestion.25 Application-agnostic network management coupled with user-controlled prioritization gives network providers the tools they need to maintain the quality of the Internet experience for all users, even during times of congestion, while preserving the application-blindness of the network and the principle of user choice, with beneficial economic, social, cultural and political consequences. Since network providers can allocate bandwidth among users using application- Network providers would be able to enforce fairness among users, allocating bandwidth among users in application-agnostic ways, but how a user decides to use its ?share? of bandwidth, both in general and at a particular point in time would be decided by the user. To the extent that applications benefit from relative prioritization at times of congestion, network providers could allow users to choose which applications to prioritize within the user?s bandwidth envelope during times of congestion. As long as the ability to prioritize is offered independently of the specific applications or classes of applications (i.e. not tied or restricted to specific applications or classes of applications) and the choice of which applications to prioritize is left to the user, this form of network management would be consistent with the non- discrimination rule proposed above. 25 For a longer explanation of the policy arguments in this subsection, see van Schewick (2008a), pp. 4-8 and van Schewick (2008b). van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 16 agnostic criteria, they can prevent aggressive users from overwhelming the network. But how users use the bandwidth available to them, and whether they would like to give some of their applications priority over others, would be choices left to the users. At the same time, the exception provides a safety valve that allows network providers to react in more application- specific ways if a problem cannot be solved in an application-agnostic way. From a technical perspective, application-agnostic network management has the added advantage of removing the incentive for users to masquerade their applications to evade or take advantage of certain application-specific treatment in the network, freeing resources at the network provider and at users. Tools for application-agnostic congestion management are available today. As the experience of Comcast shows, it is possible to protect the quality of the Internet experience of all Internet service customers in application-agnostic ways.26 The proposed rule is also compatible with new standards that are currently being developed in the Internet Engineering Task Force. Beyond Comcast?s approach, vendors have developed network management solutions that allow the network provider to allocate bandwidth among users in application-agnostic ways, while letting users choose the relative priority of applications within the bandwidth allocated to them. 27 Problems that cannot be solved in application-agnostic ways (even if coupled with user- controlled prioritization) would be captured by the reasonable network management exception. This exception would allow deviations from application-agnostic network management (and user-controlled prioritization) if the problem cannot be addressed in this way. These standards would evolve the Internet standards in a way that allows the network provider to determine how much a specific user is contributing to congestion at any point in time. This information would allow network providers to manage their networks based on a user?s contribution to congestion ? an application-agnostic criterion. Certainty and Costs of Regulation Since it is easy to determine whether a certain differential treatment is application-agnostic or not, the proposed rule clearly distinguishes acceptable from inacceptable behavior. As a result, the rule provides much-needed certainty to industry participants (including network providers, application developers and their investors) and keeps the cost of regulation low. Other Approaches 26 Bastian, et al. (2010). 27 These standards are being developed by the Congestion Exposure Working Group. See Internet Engineering Task Force (2010). van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 17 Ban discrimination that is anticompetitive or harms users I sometimes hear proposals for a non-discrimination rule that would ban discrimination that is anticompetitive and harms users. Such a rule may define certain behaviors as presumptively allowed or not allowed under these rules (e.g. user-controlled prioritization presumptively o.k.; application provider-paid prioritization presumptively not o.k.). Whether these conditions (anticompetitive, harm to users) are met and whether the presumptions should not apply to the behavior under consideration would be decided by the Federal Communications Commission (FCC) in case-by-case adjudication. The proposal for a legislative framework on network neutrality put forward by Google and Verizon in August 2010 constitutes an example of such a non-discrimination rule: ?Non-Discrimination Requirement: In providing broadband Internet access service, a provider would be prohibited from engaging in undue discrimination against any lawful Internet content, application, or service in a manner that causes meaningful harm to competition or to users. Prioritization of Internet traffic would be presumed inconsistent with the non-discrimination standard, but the presumption could be rebutted.?28 Such a rule sounds good, but does not have the desired effect. That?s because the substantive criteria do not capture the behavior that network neutrality proponents are concerned about. Beyond that, agreeing on a rule that leaves all decisions about the legality of discrimination to case-by-case adjudication by future FCCs irrevocably tilts the playing field in favor of network providers. In combination with substantive criteria that do adequately protect values such as application innovation, user choice or the Internet?s ability to serve as a platform for democratic discourse in the future, this will drastically reduce the likelihood that anybody will be able to successfully challenge discriminatory behavior in the future. Substantive problems with the criteria ?anticompetitive?: This criterion is intuitively appealing: it resonates with the notion that network providers may have an incentive to exclude applications that compete with their own applications. However, prohibiting only discrimination that is anticompetitive would only capture a subset of the cases in which network providers have an incentive to exclude applications.29 First, discrimination designed to exclude unwanted content or manage bandwidth on a network may often lack an anticompetitive motivation. In the examples of content-based discrimination that are often mentioned in the debate (e.g. Telus/Voices for Change; Verizon 28 Google & Verizon (2010), p.1. 29 The following three paragraphs are based on van Schewick (2009), pp. 36-37. van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 18 Wireless/NARAL Pro Choice),30 Second, even blocking that hurts a competitor is not necessarily prohibited by such a rule. Those who propose banning only ?anticompetitive? discrimination import the term and its meaning from US antitrust law. none of the content providers whose content was blocked was competing with the network provider. Similarly, a network provider may have an incentive to exclude or slow down selected bandwidth-intensive applications to manage bandwidth on its network, even if the network provider does not offer a competing application itself. At the same time, the resulting harm ? users? inability to participate in social, cultural or democratic discourse related to the blocked content, their inability to use the Internet in the way that is most valuable to them, or application developers? difficulty to obtain funding for an application ? is caused by the blocking as such, not by the motivations that were driving it. 31 In US antitrust law, however, the term ?anticompetitive? has a much narrower meaning than non-lawyers would expect: In particular, behavior that ?harms a competitor? (e.g. by excluding the competitor from the market) is not necessarily anticompetitive. To be anticompetitive, the behavior needs to ?harm competition.?32 For example, if a network provider excludes an application such as BitTorrent from access to the provider?s Internet service customers, this only constitutes ?anticompetitive? conduct under US antitrust law if it creates a ?dangerous probability of success? that the network provider will monopolize the nationwide market for BitTorrent-like applications. That the network provider?s customers cannot use BitTorrent, or that BitTorrent is excluded from a part of the nationwide market, is irrelevant in the context of antitrust law, but not in the context of the network neutrality debate that focuses on different types of harm. Also, US antitrust law usually has very stringent requirements about the degree of market power in the primary market that is required in order for exclusionary conduct to be problematic. By contrast, network neutrality proponents are usually concerned about any blocking by network providers, even if the network provider in question does not have a dominant position in the local or nationwide market for Internet services.33 Prohibiting only ?anticompetitive? conduct will not prevent all relevant discrimination. To protect user choice and the Internet?s ability to realize its economic, political, social and cultural potential, we need rules that prohibit blocking and discrimination of applications and content regardless of the underlying motivation and independent of the network provider?s market share. ?harm to users:? This criterion is also intuitively appealing: it resonates with the notion that network neutrality is designed to safeguard users? ability to use the applications and access the content of their choice without interference from network providers. However, this criterion may 30 For a description of these examples, see van Schewick (2010a), pp. 266-269. 31 For one example, see Farber, et al. (2007). 32 See Farber, et al. (2007). 33 This concern is driven by the insight that a network provider may have an incentive to exclude applications even if it competes with other network providers in the market for Internet services. See van Schewick (2010a), pp. 255- 264. van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 19 not capture important instances of discrimination that network neutrality proponents are concerned about. Consider the example of Comcast?s blocking of BitTorrent. Network neutrality proponents usually agree that singling out specific applications to manage bandwidth on a network is not an acceptable form of discrimination (aka ?reasonable network management?) as long as other, application-agnostic ways of managing the network are available. Now imagine how the proposed rule would apply to this case. The rule immediately raises lots of questions: Who is a user? Only end users, or also application and content providers? How do regulators determine whether a user is harmed? Do they focus on the individual user who cannot use the Internet as she would like, or do they focus on users as a group (similar to the way antitrust law defines harm to consumers when evaluating whether a certain conduct is anticompetitive)? For example, slowing down peer-to-peer file-sharing, a network provider may argue, may harm the file-sharing users and the provider of the file-sharing software, but, according to the network provider, is only done to protect the Internet experience of all the other innocent non-file-sharing users. Does it matter that there are alternative, non-discriminatory ways of managing the network that are not similarly harmful to the users of the file-sharing software and the providers of the software, while maintaining the quality of the Internet experience for the non-file-sharing users? I don?t think it matters under this formulation of the rule, but network neutrality proponents usually think it should. What about the fact that peer-to-peer file-sharing is often used by individual filmmakers to inexpensively distribute their creative works, as we know from the Canadian proceeding that reviewed the Internet traffic management practices of Internet service providers? What about the fact that peer-to-peer file-sharing is often used by non-profits to distribute their video contributions to political debates? How would regulators integrate consideration of these societal benefits (fostering a more decentralized environment for democratic discourse and cultural production in which anybody can participate) into the assessment of the discriminatory behavior? Again, these considerations should matter, but are probably not captured by this formulation of the rule. ?anticompetitive and harm to users:? It is also unclear whether such a rule would require one or both criteria to be met in order for discrimination to be considered harmful. This may depend on the exact formulation of the rule ? whether the rule bans discrimination that is ?anticompetitive and harms users? or discrimination that is ?anticompetitive or harms users.?34 34 The Google-Verizon Proposal for a legislative framework for network neutrality said ?or.? See the wording of the proposed non-discrimination rule on p. 2 of this paper. If the rule says van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 20 ?and,? both criteria need to be met. Given the difficulties of showing that behavior is ?anticompetitive,? the second criterion becomes effectively useless. In sum, the rule?s substantive criteria are open to interpretation. They do not accurately separate socially beneficial from socially harmful discrimination. In particular, the rule would not prohibit many instances of discrimination that network neutrality proponents are concerned about, leaving application developers and users without protection. The rule makes it impossible to consider the potential impact of discriminatory conduct on the Internet?s ability to realize its social, cultural and political potential ? important aspects that the network neutrality rules are intended to protect. Problems with Case-by-Case Adjudication The rule uses criteria that are open to interpretation. It does not indicate which economic theories of competition and user harm should guide the assessment of discriminatory behavior. Instead, it leaves all substantive judgments about the legality of discriminatory behavior to later adjudications. This may be appealing to parties looking for a compromise today, since the controversial questions (?Which forms of discrimination should be banned??) are not decided one way or the other. From the perspective of the regulator, this strategy suffers from some important flaws: First, it fails to provide much-needed certainty for industry participants. Network providers still won?t know which forms of network management are acceptable, which constrains the evolution of the network more than necessary. Application developers still face the fundamental risk that the network will turn against their application at any time and block the application or slow it down, which will reduce their ability to get funding. At the same time, the existence of the non- discrimination rule provides little comfort to application developers who consider whether to realize their innovative idea or to investors who consider whether to fund them. After all, venture capitalists and other investors fund start ups so that they can build their product and better meet the needs of their users. Paying armies of lawyers and economists to clarify how to interpret an ambiguous non- discrimination rule in order to allow the application to reach its customers is not how they would like their funds to be used. Second, it creates high costs of regulation and tilts the playing field against those ? end users, application developers and start-ups ? who do not have the resources necessary to engage in extended fights over the legality of specific discriminations in the future. Stakeholders cannot agree on what constitutes acceptable behavior today; it is unlikely that they will be able to do so in the future. Still, the strategy is appealing because each party can hope that it will prevail in the future. An ambiguous rule that leaves everything up for debate fits well with that goal (the Telecommunications Act of 1996 is a good example of that strategy in action). But for the regulator, who is charged with protecting the public interest, other van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 21 considerations should matter. With an ambiguous rule like this that is applied in case-by-case adjudication, the party that can pay the most lobbyists, lawyers and economists wins. End users, low cost application developers and start-ups loose. They do not have the resources to pay for the lawyers and economists needed to win the fight over the correct interpretation and the application of the rule, first at the regulatory agency and later in the courts. They are, however, some of the key groups that network neutrality rules are intended to protect.35 Finally, insights from behavioral economics suggest that discriminatory behavior is more likely to be allowed under case-by-case adjudication than under an ex ante regime. An ex ante regime is better suited to the consideration of the very fundamental values at stake. Network neutrality rules are based on very general trade-offs among competing values.36 ? that fostering application innovation is critical for economic growth; They are based on the view ? that allowing users to choose how they want to use the network is important if we want to maximize the Internet?s value for users, and for society; ? that we need to preserve the Internet?s ability to foster democratic discourse and so on. These goals are considered more important than ? the potential benefits of allowing the core of the network to evolve freely; ? the potential benefits of short-term optimization of the network in favor of the applications of the day; and ? network providers? profits. While the benefits of the network neutrality rules are difficult to specify in detail (they will be realized in the future; we don?t really know the applications which we will never get if we close down the network today; we don?t really know which benefits a better platform for social, cultural and political interaction will bring), the costs are immediately apparent (?we would like to do X (i.e. manage bandwidth this way) and can?t do it?).37 Insights from behavioral economics suggest that decision makers tend to undervalue future benefits, even more so if they are uncertain. In addition, for the decision maker, allowing one deviation from a general non- discrimination rule does not seem to make a big difference. This assessment may be mistaken; often, it may be impossible to immediately recognize the negative future consequences. Finally, on the Internet, many small deviations quickly add up to create big roadblocks for innovation.38 35 On the importance of low cost innovators, see van Schewick (2010a), pp. 204-213, 300-308, 334-345. On new entrants and start ups, see van Schewick (2010a), pp. 319-334. For all these reasons, deciding whether to allow discrimination on a case-by-case basis makes it 36 For a detailed discussion of this trade-off, see van Schewick (2010a), pp. 355-371. 37 For a more detailed description of the problem with pointers to the literature, see van Schewick (2010a), pp. 77- 78, 374-375. 38 For an example, see van Schewick (2010a), p. 78. van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 22 more likely that discrimination will be allowed than under an ex ante rule that resolves the above trade-off for all future cases at once. Ban discrimination that is not disclosed This is the approach adopted by the European Union as the result of the review of its regulatory framework for telecommunications services. This is based on the idea that if a network provider discriminates against an application, content or service that users would like to use, users can switch to another network provider who does not discriminate against the affected application. The threat of switching, proponents of this approach assume, will then discipline providers. Disclosure can only facilitate competition and discipline providers if there is effective competition. In order for disclosure to have a disciplining effect, customers need to be able to switch to another provider that does not impose a similar restriction, and they need to be able to do so at low costs. As I have explored elsewhere, these conditions will often not be met, making this rule an ineffective safeguard against discrimination. This approach will be discussed in more detail in later versions of this white paper. REFERENCES Bastian, C., T. Klieber, J. Livingood, J. Mills & R. Woundy. 2010. "Comcast's Protocol- Agnostic Congestion Management System." Internet Draft IETF. Briscoe, Bob, Toby Moncaster & Louise Burness. 2007. "Problem Statement: We Don't Have To Do Fairness Ourselves " Internet Draft IEFT. Crawford, Susan C. 2007. "The Internet and the Project of Communications Law." UCLA Law Review, 55(2): 359-407. Farber, David, Michael L. Katz, Gerald Faulhaber & Christopher S. Yoo. 2007. "Hold Off on Network Neutrality." Washington Post. January 19. http://www.washingtonpost.com/wp- dyn/content/article/2007/01/18/AR2007011801508.html Google, Inc. & Inc. Verizon. 2010. "Google-Verizon Legislative Framework Proposal." August 9. http://static.googleusercontent.com/external_content/untrusted_dlcp/www.google.com/en/us/goo gleblogs/pdfs/verizon_google_legislative_framework_proposal_081010.pdf Internet Engineering Task Force. 2010. "Congestion Exposure (conex) Working Group Charter." https://datatracker.ietf.org/wg/conex/charter/ Internet Non-Discrimination Act of 2006. S. 2360. 109th Congress, 2nd Session. 2006. van Schewick, Barbara. 2008. Official Testimony at the Federal Communications Commission Second En Banc Hearing on Broadband Management Practices. Federal Communications Commission. van Schewick ? Network Neutrality: What a Non-Discrimination Rule Should Look Like Center for Internet and Society White Paper ? Version 1.0, December 14, 2010 23 van Schewick, Barbara. 2008a. Official Testimony at the Federal Communications Commission Second En Banc Hearing on Broadband Management Practices. Federal Communications Commission. van Schewick, Barbara. 2008b. Oral Testimony at the Federal Communications Commission Second En Banc Hearing on Broadband Management Practices. Federal Communications Commission. van Schewick, Barbara. 2009. "Point/Counterpoint. Network Neutrality Nuances." Communications of the ACM, 52(2): 31-34, 36-37. van Schewick, Barbara. 2010a. Internet Architecture and Innovation. Cambridge, MA: MIT Press. van Schewick, Barbara. 2010b. Opening Statement at the Workshop on Approaches to Preserving the Open Internet. Federal Communications Commission. van Schewick, Barbara. 2010c. Opening Statement at the Workshop on Innovation, Investment and the Open Internet. Federal Communications Commission. Wu, Tim. 2006a. "The AT&T Network Neutrality Agreement." December 29. http://www.timwu.org/log/archives/81 Wu, Tim. 2006b. "Why Have a Telecommunications Law? Anti-Discrimination Norms in Communications." Journal on Telecommunications and High Technology Law, 5: 15-46.