 Well, hi there. I'm Jonathan Zittrin, and I am so honored today to be here to introduce Barbara Van Shevik, both because of what she has to say today and because of the amazing brain trust I already see in this room. So I'm very much anticipating the question and answer session that will follow. Barbara is a computer scientist who is also a law professor, sort of rarity and a wonderful sort of CP snows clashing of cultures. Barbara makes the most of it, and it has been one of the leading voices and scholars looking at the architecture of the internet, the way in which it connects to innovation, the threats that it faces, implausible ways around them using both technology and law as a guide. A lot of the arguments around that neutrality and the architecture of the internet, some of the arguments have been this is all theoretical. There's no reason to call a fire department yet. Nothing yet is happening for which we should be all that panicked. Barbara was one of a handful of people who actually saw a fire burning in Comcast, manipulating bit torrent traffic, and was one of those who petitioned the FCC to do something about it in a situation in which the FCC did. It went up to the courts. You may hear about that today. But in other words, Barbara has been somebody really very much, not just working on the theory of this, but also on the practical implications. It's 1215 right now. We don't have a whole lot of time. We'll end no earlier than 115, no later than 130, in revenge of the people who sat in here longer than they do before noon. And that means we should get right to it. But let me say that it's so great to have one of our counterparts from Stanford here. We hope she'll stay. The weather is not promising on that front. But in the meantime, Barbara, thank you so much for coming today and giving a talk for which we understand this is only the second time you have delivered this talk. So we feel very pleased and privileged to be guinea pigs for you. Thank you very much. Yeah, thanks a lot, Jonathan, for your kind introduction. And thank you all for coming. Being here is always a pleasure and, of course, just looking around and seeing so many friends and colleagues in the room is great. So in the past 30 to 40 years, the internet has created an enormous amount of application innovation. In my book, I argue that this is not an accident, but a result of the internet's internal structure. Or as network engineers call it, it's architecture. Today, the architecture of the internet is changing. And some of these changes are driven by network providers' desire for more profits. Other changes are the result to technical challenges that the internet is facing today. But all this raises an important question. What exactly are the factors that have fostered application innovation in the past? Because these factors are the factors that need to be preserved if we want the internet to serve as an engine of innovation in the future. I want to talk about this through three stories and here's the first. In the mid-1990s, a software engineer in Silicon Valley named Pierre Armandier thought, wouldn't it be nice if we could buy and sell things over the internet using auctions to determine the price? Most of his friends, he was totally crazy. Strangers will never buy from strangers online, they said. But Pierre Armandier didn't care about this. He stayed home over Labor Day weekend, 1995, wrote his software and put it online. Nine months later, so many people were using the software that he decided to quit his day job and focus on the platform online. The platform renamed eBay became a huge success and today more than 88 million people worldwide are using eBay to buy and sell things over the internet. Three aspects are important here. So Pierre Armandier had an idea for an application. People thought it was crazy, but it didn't matter. In particular, it didn't matter whether network providers thought that strangers would ever buy from strangers online. And that's because on the internet, network providers don't have to do anything to enable a new application to run. And that's the consequence of the internet's architecture. The original internet was based on a design principle called the end-to-end arguments. This design principle was first identified and described by Jerome Salzer, David Reed, and David Clark, and David Reed is actually here today, sitting right next to Jochai, so I'm glad you're here. As I show in my book, there are actually two versions of the end-to-end arguments, which has created a lot of confusion in the past. And here I'm talking about the broad version of the end-to-end arguments. Following this design principle, the network was designed to be as general as possible in order to provide and support a wide variety of applications with different needs. So when a new application comes along, the network doesn't have to be changed to enable the new application to run. All the innovator has to do is write a program that runs on a computer attached to the internet. As a result, an innovator does not have to convince network providers that her application is useful or will be commercially successful. The only person who needs to be convinced that this is a good idea is the innovator herself. This greatly increases the chance that innovative ideas will be realized. Second, when the application has been written, the network doesn't need to be changed before the application can actually run on the network. If you want to use it, you install it on your computer. That's it. Again, the only person who needs to be convinced that this application may be useful is the person who actually wants to use it. This greatly increases the chance that people can actually use the application. And third, in this architecture, it doesn't cost a lot to develop new applications. You need access to computer, be able to program, and time to actually write the program. This greatly increases the number and type of people who can develop new applications. Like Pierre Armadier, you don't have to be an employee of a firm or have outside funding to realize your idea for an application because the biggest investment is often that time spent designing and programming the application itself. Innovators can develop an application in their free time or as a site project. And under these conditions, an application doesn't have to produce a profit in the future to cover the cost of developing it. Instead, a wide range of benefits, such as, you know, fun programming or the wish to use this application may be sufficient to cover the development cost. And this means the architecture allows innovators with a wide range of motivations and funding models to develop new applications. So three aspects. Innovators independently decide whether to realize innovative ideas. They do not need support or permission from network providers in order to innovate. Users independently decide whether they want to use an application. And the low cost of application innovation enable a very large and diverse group of people to develop new applications. This was the first story, here's the second. In 2002, two European entrepreneurs named Nicholas Zendstrom and Janus Freistad, wouldn't it be nice if we could use peer-to-peer software to make phone calls over the internet? At the time, most network engineers thought this wouldn't work. They thought internet left me required special treatment from the network, something we call quality of service. Network providers weren't really interested in pursuing internet telephony because the application constituted a huge threat to their business model. But Zendstrom and Freist didn't care about this. They went ahead, developed their software, the software became Skype, and today more than 500 million people worldwide are using Skype to make phone calls over the internet. Again, we have two entrepreneurs who have an idea for an application. Network engineers didn't think it would work, but it didn't matter. The application constituted a huge threat to network providers' business models, but it didn't matter. And for Zendstrom and Freist, it didn't matter because there was nothing network providers could do about this. And there was nothing they could do about it because the architecture of the internet prevented network providers from interfering with the applications and content on their network. As I said, the internet was based on the broad version of the end-to-end arguments. And as a consequence of this design, the network couldn't distinguish between the different applications on its network. It was application blind. And as a result, the network providers couldn't control the applications on their network. Today, that's different. Today, sophisticated technology is available that enables network providers to identify the applications and content on their networks and control their execution. Thus, the original internet was application blind. Today's internet is not. Does it matter? Imagine you have this great idea for a video platform that will revolutionize the way people watch TV. Once they've used your application, they will never go back to cable again. Of course, there are risks. The technology may not work. Users may not like your product. Your business model may be wrong. But in the application blind network, you know that you'll have a fair chance in the marketplace. You'll be able to compete with other applications on the merits. In today's network, network providers may squash you. The network can turn against you any time and block your application or slow it down. There are many reasons why network providers may want to do so. Maybe your application competes with theirs, like Vonage competed with Madison River's traditional telephone service, or they just want to share of your profits. Maybe they don't like your content the way TALUS, the Canadian network provider, didn't like the discussions that were happening on the website of the union member with which TALUS was in strike. And so just block the website of that union member. All your applications slow down to manage bandwidth on a network the way the torrent was slowed down because Comcast wanted to manage bandwidth. But whatever the network provider's reasons, if your application gets blocked, your project fails and you won't be able to reap its benefits. And accounting for this possibility, you, or potential investors, may decide not to pursue your idea. Third story. When YouTube came on the market, it competed with Google Video. YouTube was better. It won. In today's internet, things might have been different. In an application-aware network, network providers can ask application providers to pay an access fee, a fee for the right to reach the internet service customers of the network provider. There are many ways in which a network provider could do so and I'm happy to talk more about them in the discussion. Now I wanna focus on one possibility. When YouTube came on the market, network providers might have said, Google, you are rich. You have lots of money. Why don't you give us some of that money and we'll give Google Video better transport? Imagine Google pays. Suddenly, Google Video is so much better. Not because it's a better product, but because Google is rich and Google was able to strike a deal. Suddenly, in such a world, network providers get to decide which applications are successful by deciding who gets the deal. Now, network providers not users get to decide and to pick winners and losers on the internet. And I'll talk about why this is a problem in a second. But as soon for a moment, we decided to present this problem by forcing network providers to offer in charge for these services in non-discriminatory ways. So if AT&T offers better transport to Google Video, it also needs to offer it to YouTube. But even then, these types of access charges would still give an advantage to established companies with deep pockets over innovators with no a little outside funding. Why? How would innovators with no a little outside funding pay the access fee? After all, they don't have a lot of money. Subscription fees? Existing applications usually don't have subscription fees. Advertising revenues? New applications usually don't command significant advertising revenue. Venture capital might be a solution, but we know that innovators are often unable to get it. Thus, many innovators would not be able to pay for access charges, giving those who can pay an immediate competitive advantage. And for seeing this, innovators with no a little outside funding may stop innovating on quality of service sensitive applications altogether. So three stories, three different factors. How do changes in these factors affect the amount and quality of application innovation? Some changes may affect the cost and benefit of application innovation. An innovator decides to innovate if the benefits broadly defined are larger than the costs. Increase the costs or reduce the expected benefits and some applications may not be justified anymore. And you know, you might reduce the benefits, for example, by charging access charges or discriminating against the application. Other changes may affect the size and diversity of the innovator pool. For example, as I said, access charges may affect all application developers, but they may hit certain types of innovators, those with no a little outside funding, particularly hard. And finally, other changes may let network providers and not users decide which applications will be successful and how the network can be used. Why are these things important? If there is uncertainty or if user needs are very heterogeneous, a larger and more diverse group of innovators will develop more and better applications than a smaller, less diverse group of innovators, and these applications will better meet user needs. What's the intuition? If there is uncertainty, nobody really knows in advance which applications will work or will be successful. Under these circumstances, economic theory suggests that it's best to try out many different ideas and see what happens. Some will succeed, some will fail, but trying is the only way to find out. And because different people will have different ideas and different views of the world, more and more diverse people will have more and more diverse ideas. A larger and more diverse group of innovators will also realize more of the ideas that are known. For example, startups may have an incentive to realize innovative ideas that established firms may not find attractive or users who want to use the applications themselves may have an incentive to create an application that a manufacturer who wants to sell this application would never produce. By contrast, fewer innovators or less diverse innovators will try fewer things, leaving valuable ideas on the table. So innovative diversity is important, but that's not the whole story. My book highlights the importance of a specific subgroup of innovators, those with no or little outside funding. As I said in the original architecture of the internet, almost anybody can be an innovator. This allows innovators with a wide range of motivations and funding models to develop new applications. By contrast, things like access charges would remove the ability of a large subgroup of innovators, those with little or no outside funding to develop new applications. The consequences of this loss would be severe. Throughout the history of the internet, many of the most important innovations have been developed by innovators of this type. eBay or Facebook, which were developed by innovators in their free time. Google or Yahoo, which were developed by grad students. The Apache web server or the World Wide Web, which were developed by users, or Flickr or Blogger, which were developed by startups with little or no outside funding. Often, these innovators try to get venture capital funding, but weren't successful. This happened, for example, to eBay, Google, Blogger, and Flickr. So in a world where network providers can charge access charges, these applications wouldn't have been able to get off the ground. I sometimes hear that the internet has matured. So many applications are now, application providers are now multimillion-dollar companies. So maybe we don't need low-cost innovators anymore. But I don't think that's correct. There are many more applications out there waiting to be developed. And as much as I like Google or Amazon or Yahoo, I don't think they'd be able or even interested in developing many of them. Think about it. Classifieds were revolutionized by Craigslist and eBay, not by the newspapers. Music distribution was revolutionized by Napster, Kaza, Bittorrent, or iTunes, not by the music industry. The commercial potential of the World Wide Web was realized by Netscape, not by Microsoft. And online bookstores got off the ground by Amazon, or with Amazon, which was founded by former senior vice president at a quantitative hedge fund with a degree in computer science and electrical engineering, not by bonds and noble. My research suggests that this is not an accident. Established companies have very different backgrounds, cost structures, and motivations that let them focus on very different applications than the types of applications that the different types of low-cost innovators would focus on. I don't want to be stuck with the applications that these established companies chose to develop. Thus, if we want to protect application innovation, we need widespread experimentation by a large diverse group of innovators, and this group should include innovators with no or little outside funding. But widespread experimentation is only part of the mechanism that produces innovation under uncertainty. The second is who gets to decide which applications become successful? Users or network providers? Does it make a difference? I argue it does because network providers and users will choose different applications. There are two reasons for this. First, users and network providers use different criteria when choosing which applications will be successful. Users choose the application that best meets their needs. That's easy. Network providers may choose or may use different criteria. Does this application compete with my own application? Does it create lots of bandwidth? Does my preferred vendor offer network management software that happens to block this application? Consider Skype. Many mobile providers in Europe do not allow their users to use Skype on their cell phone over the mobile internet. If you look at user forums, you'll see that users don't like that. They want to use Skype on their cell phone. But if users use Skype, they don't make as many traditional cell phone calls and voice revenue shrinks. So network providers make a decision that's different from the decision that users would make if they have the choice. And second, even in those cases where network providers would actually like to choose the application that users want, they don't necessarily know what that is. That's the uncertainty I talked about earlier. In many cases, nobody knows whether an application will be successful until users actually try it. And network providers just can't replace this. Beyond innovation, user choice is also important if we want the internet to provide maximum value to society, but that's a different story. So if network providers pick winners and losers on the internet, if they decide how we can use the network, we may end up with applications that we wouldn't have chosen and may be forced to use the internet in a way that doesn't create the value it could. In some, there are a number of aspects that foster application innovation. Innovators independently choose which applications they want to use. Users independently choose which applications they want to use. And innovators, of course, choose the applications they want to develop. The application blindness of the network ensures that the network provider cannot interfere with these choices, that it cannot distort competition among applications or classes of applications, and that it cannot reduce application developers' profits through access charges. Finally, the low cost of innovation not only make many more applications worth pursuing, but also allow a very large and diverse group of people to develop new applications which in turn increases the overall amount and quality of innovation. Allowing network providers to discriminate or charge access charges breaks this mechanism at its core. It reduces application innovation. It reduces the value users and society can realize from the internet. And it also, though I didn't have time to talk about this, removes the very features that have allowed the internet to improve democratic discourse, to facilitate democratic political organization and action, and to create a decentralized environment for political and social interaction and cultural interaction in which anybody can participate. Thus, blocking discrimination or access charges are not in the interest of society. Network providers' interests point in a different direction. Blocking discrimination and access charges increase their profits. At the same time, network providers capture only a very small part of the social benefits associated with things like application innovation, free speech, or a more participatory culture. And even the gains they are able to capture are uncertain and mostly in the future. So network providers consider them even less. However, network providers control the network infrastructure and so it's not surprising that they change this network infrastructure in order to favor their commercial interests. Thus, there is a gap between network providers' private interests and the public interests, or as lawyers and economists call it, a market failure. Given this gap, we face an important choice. Leaving the evolution of the network to network providers will significantly reduce the internet's value to society. If we want to preserve this value, policy makers will have to intervene and preserve the factors that were at the core of the internet's success. And that's exactly what the network neutrality debate is all about, protecting the features that were at the core of the internet's success. How to do so without overly constraining the evolution of the network or without increasing high cost of regulation is a difficult question. And I hope we'll be able to talk more about that in the discussion. But I want to be very clear here because I'm sometimes misunderstood on this topic. I'm not saying we need to preserve the original architecture of the internet. I'm saying we need to preserve the factors that have fostered application innovation in the past and made the internet valuable for society. So we need to preserve the factors that I've just talked about. And that's not the same as preserving the original architecture of the internet. And because that sounds totally abstract, I want to talk for a moment about the relationship between network neutrality and quality of service so that you can see what I mean. So what's quality of service? In the original internet, the internet provides a single best effort service, which means the network does its best to deliver data packets, but it doesn't provide any guarantees with respect to delay or bandwidth or losses. And in this sense, it's very much like the normal vision of the postal service. And contrary to the postal service, where you can also choose two-day shipping or other forms of expedited shipping, the internet, the original internet only offered this one single service to all data packets. But different applications have different needs. So for example, internet telephony is really sensitive to delay, but doesn't care about the occasional packet loss. Whereas email cares a lot about packet loss that doesn't really care about a small amount of delay. And so we could imagine a network that treats different applications differently, according to their needs. So it could give low delay to internet telephony and then just the normal best effort service to email. And this type of thing where the network offers different types of service to different data packets is what we call quality of service. In the network neutrality debate, this question comes up when we ask how we should define discrimination. And I just wanna give you a tiny bit of background if you're not following the network neutrality debate on a day-to-day basis where this question comes in. So if I'm a regulator and I wanna consider whether we need network neutrality rules in my country, I need to ask myself a series of questions in order to decide what these rules if we need them should look like. And the first question is, do we need a rule against blocking? That's the story of Skype. And so most network neutrality proponents or all network neutrality proponents think there should be a rule against blocking. So you could even say, if there is no rule against blocking in a proposal, it's not a network neutrality proposal. That's the one defining factor that holds all network neutrality proposals together. But what about behavior that stops short of blocking? So for example, if Comcast slows down internet video applications like YouTube or Hulu, that compete with Comcast's own internet video application, fancast, should this be prohibited? Within the network neutrality debate, the term for this type of behavior is discrimination. And so the question is, should we ban only blocking or also discrimination? And if we decide to ban blocking, how should we define this term? So the first option, and this is the one that network providers favor most, if they accept at all that we might need network neutrality rules, is that they say quality of service would be really beneficial. Just think of all these amazing applications that we'll never get because we don't have quality of service. And then they say, but it's really difficult to distinguish beneficial discrimination like quality of service from harmful discrimination and we don't really think harmful discrimination exists anyways. So we should just allow all forms of discrimination, which is effectively equal to having no rule against discrimination. So in such a regime, you would only have a rule against blocking. And I don't think that's a good solution because it goes too far because just banning blocking but allowing discrimination makes the rule against blocking meaningless. And that's because by slowing down an application, you can make it effectively unusable. And at the same time, users may not realize what you're doing. So you have leave open an option that's just as effective but actually less costly. And that makes the rule against blocking meaningless. So that's not the right choice. And other people say, and Susan Crawford for example would be an example, that we need to ban all discrimination. We need to require network providers to treat every packet the same. And these people often point to the original internet or they often say that network providers have a really network engineers are divided about whether quality of service is actually a useful and feasible technical solution. And many network engineers think it would be much better to just increase capacity. And so people who advocate for this approach also point to these technical debates and to the fact that internet too for example has publicly said, we think quality of service is really too complicated and let's just go with increasing capacity. But I don't think that's the right answer either because it's too restrictive. You know, yes, we don't want network providers to distort competition among applications or classes of applications, but if we could do that without totally constraining the evolution of the network, that would be much better. And to the extent there is a technical debate about the value of quality of service, I really don't think it's the job of regulators to determine these technical debates. We should only constrain the evolution of the network to the extent it's absolutely necessary to protect the values that we want to protect. And I'll show you that this is possible. So the all on nothing approaches don't work. And my favorite approach is one that tries to reserve the application blindness of the network. So, you know, you can see this the first factor that we identified as one that fostered application innovation in the past. And we could do this in two ways. So we would want to bend our application specific discrimination, but everything that's application agnostic would be fine because that's the same as in an application blind network. Their network providers could engage in application agnostic ways of managing the network. For example, you know, managing bandwidth, just based on the amount of bandwidth that users use. So what is application specific discrimination then? And there are two different options. One would be to say, we only ban discrimination that's based on specific applications. So you cannot take a single application and excluded or slow it down. So Comcast can't take YouTube and slow it down. The other option is we ban all discrimination based on applications or classes of applications. So you can't even treat different classes of applications differently. So let's start with the first. If you only ban discrimination based on application, then network providers wouldn't be able to treat fancast better than Hulu, but also they wouldn't be able to treat YouTube better than Hulu. So it doesn't matter whether the application that's favored is the network providers application or not, but they would be allowed to discriminate based on the class of application. And that's something that has been suggested as like treatment. So basically network providers would be allowed to treat or would be required to treat like applications but they could distinguish between applications with different needs. So in practice, this would mean that for example, you could treat internet telephony different from email because they have different characteristics and different needs, but you couldn't treat Skype differently from Vonage because they would be members of the same class of applications. And the idea here is that it doesn't really hurt email if internet telephony gets lower delay because email doesn't really care about delay anyways. And this type of rule for example, was included in the Google Verizon proposal as part of their network management exceptions where they said like treatment is generally fine as reasonable network management. And sounds really innocent, but as you see, I don't think that's a good solution and that's because there are four problems. The first is if you require like treatment, it's not really obvious what the classes should be. And this problem with defining classes gives network providers the ability to distort competition among applications. And I'll give you two examples. The first is they could use that to distort competition among classes of applications. So they might say online gaming needs low delay and selling low delay to online gamers would be quite attractive because it would enable network providers to capture some of the values that online gamers get from gaming. So network providers might like that. But then they might say, we don't really wanna give low delay service to internet telephony because that would enable internet telephony services like Vonage or Skype to compete better with Comcast Digital Place and we don't really want that. So we say one class is online gaming, we give it low delay service, the other class is internet telephony, we don't give it low delay service. Now all the network engineers in the room could say, but that's the wrong way to define the classes because they are both sensitive to delay. So the real class should be delay sensitive applications. But that's a real fight that you can have. And until you succeed in having this fight in front of the FCC, the network provider can use this uncertainty to gain the system and distort competition among classes of application here, internet telephony versus online gaming. But they could even use that to distort competition among applications. And the Canadian network management proceeding provided an interesting example of that type of behavior. So in the Canadian network management proceeding, we found out that a lot of the Canadian network providers slow down peer-to-peer file sharing applications either all the time or during times of congestion. And then they said that really helps real-time applications like YouTube, that stream video in real-time to perform better under congestion. Well, then this application came along that was called VOOS. And what VOOS does is stream video in real-time using a peer-to-peer file sharing protocol called BitTorrent. So how do network providers in Canada treat VOOS in times of congestion? Do they say you belong to the class of peer-to-peer file sharing applications and therefore we'll slow you down and thereby they put it at a disadvantage as opposed to YouTube, which also streams video in real-time but doesn't use peer-to-peer file sharing? And we don't really know what they did, but the existence of this example shows you that even within the same class of application, you may use this sort of class-based approach to put certain applications at a disadvantage. And of course you could always go to the FCC or you regulate and complain about that, but you can immediately see that these types of debates and discussions which require lots of engineers would create quite high cost of regulation. And then there are two more problems. One is it's the network provider, not the user who decides which applications get quality of service. And that conflicts with this idea of user choice and the importance of user choice that I pointed out earlier. Why is this a problem? This doesn't really meet user preferences that well because user preferences aren't actually fixed over time. So when I'm at home and I'm waiting for a video to download and while I'm doing that, I'm just chatting on the phone with my friend, I may not care that much about the quality of the voice conversation. I really wanna watch that video and I wanna watch it, I want it to download really, really fast. And I contrast the next morning when I'm on a video conference or on a job interview using Skype, I want the quality to be crystal clear. And so the night I don't wanna have quality of service for my voice call the next morning, I do. But the network provider with their class-based approach with which is static will never be able to capture that difference. And so the internet will be less valuable or create less value for me and it would be if I had the choice. And then innovators need support from the network in order to get the quality of service they need. That's because the network provider defines the classes and needs to make changes in the network to make sure that the network recognize you, your new application as belonging to a certain class. And so you can immediately see how this has a number of problems. If you're a startup, you don't have this big department that's ready to go out and call network providers all over the world to ask them to allocate you to this certain class. So there are huge transaction costs. But even beyond that, once you get to the network provider, they may say you don't belong to this class and they may say that because they don't get it, which is as we've seen in the history of the internet quite likely or they might say it for more malevolent reasons because they somehow wanna hurt you also. But in any event, this is the same as the eBay story. The moment you require support from the network providers, you make it less likely that your application can actually do what it wants. And so for these four problems, I'm against the idea of like treatment, even though if you don't really think about it, it sounds quite innocent. And instead I think we should do this. We should ban all application-specific discrimination, which is discrimination based on application or class of application and only allow application agnostic discrimination. That means you cannot treat Skype different from Vonage, that would be discrimination based on application and you can't treat internet-to-left need differently from email. And that's the idea of class-based discrimination. So now you might say wait a minute, you have told us that quality of service might be really useful. And now you're saying you can discriminate or distinguish between classes of applications with different needs. That doesn't really make sense. And so that's the beauty of this proposal, that it allows certain but not all forms of quality of service. And the way this would work, and that basically immediately follows from the way that non-discrimination is defined, the network provider would be able to offer different types of quality of service. So it could offer a class that has best effort service, a class with low delay service, a class with less than best effort service or a class with guarantee bandwidth, and then users are able to choose whether and when they want to use which class of service. So I may use low delay service for online gaming. He might use it for internet-to-left need and my crazy grandmother might use low delay service for email, which doesn't make any sense, but it would be her choice. And so she could do that under this system. And then there is a third condition which doesn't really follow from the non-discrimination rule, which you would have to include separately, which is that the network provider can only charge its own internet service customers, not the application provider. And that's the story of YouTube that we talked about earlier. So we would have the ability to do quality of service, but only in this way. And suddenly all the problems that the like best treatment had disappear. So network providers can't distort competition anymore because they don't have any say in who gets which quality of service. They can only do application agnostic things. So they are not able anymore to interfere with specific applications. Users get to choose when and which quality of service to use. So these types of changes and preferences can be perfectly met, which creates higher value than like treatment. And the network provider doesn't need to support the application anymore. The only two people who need to be convinced that the application needs quality of service are the innovator who sometimes somehow needs to communicate this to the user and then the user himself or herself which just needs to signal this to the network. And so the network provider again has basically taken out of the equation. At the same time, we don't constrain the evolution of the network more than absolutely necessary to realize the values that we want to preserve. We have low cost of regulation because this criteria is pretty intuitive. You know, application agnostic against application specific and that provides much needed certainty for industry participants. So if I'm a network provider, I actually know how I can use the network. And if I'm an innovator, I have the certainty that it won't be discriminated against. So I would say we need to ban discrimination based on applications or classes of application only allow application agnostic discrimination. And the whole point of this, of course, in addition to talking about the idea of non-discrimination rules was to show that preserving the factors that have positive application innovation in the past is not the same as locking in the original architecture of the internet because it would allow us to introduce certain types of quality of service and other beneficial evolutions of the architecture of the internet. And at the same time, I hope you have seen how considering the factors that I talked about, application blindness, user choice, innovators don't need support from the network really help us think and make decisions when we evaluate alternative network neutrality proposals. So of course, these are not the only non-discrimination rules. The other one that I show first is the one that Google Verizon suggested. And the final one that's ban discrimination that's not as close as the European solution. And we can talk more about that in the discussion if you want to. But there is one final question that I want to talk about. I've talked a lot about application innovation. But why do we care so much about application innovation? Why should policy makers care about this? I have a really long answer to this question which explains how application innovation contributes to economic growth and how it creates value in all areas of society. But I really want to get to this question. So let me just say this. Have you ever tried to explain to your partner's grandmother why she should get the internet? I did. And although I'm not a computer scientist, although I am a computer scientist, I didn't say grandma, you have to get the internet. It's so cool. It lets you send data packets back and forth. No. I said, if you get the internet, you can call us and you can see your grandchildren on the screen. And if we post pictures, you can see them immediately after we post them. And you can read about everything you can possibly imagine. Thus, the internet doesn't create value through its existence alone. It creates value by enabling us to do the things we want to do, do things we never knew we wanted to do, and do things more efficiently. And applications are the tools that let us realize this value in all areas of society. And by protecting the factors that have fostered application innovation in the past, we can make sure that the internet will be even more useful and valuable in the future. Thank you for your attention. Let me just throw though something at you to start with. What do you do with Apple iPhone? That is to say several dimensions of pushback, an area where tremendous user adoption is being driven in the mobile data connectivity, driven precisely by a platform that reverses many of your assumptions while preserving a lot of relatively open applications, smack in the middle of possibly the most controversial boundary of whether or not that neutrality shooter shouldn't expand to wireless. So you've got, at one level, the pushback from should it go to wireless and at the other of, well, actually, here's an enormous space of innovation that's been enormously important and actually doesn't work like this at all. It works like the government doesn't regulate it and companies find out how to harness distributed innovation anyway. People look at the iPhone and say, look, I have a closed device where Apple controls which applications get on the iPhone and so we see all this application innovation. So aren't you totally wrong? And I would say no, because, you know, the way you need to think about this is that innovation is enabled by practice along different dimensions. And if you think of the original cell phone market, cell phones were closed devices where it was really expensive to innovate and only a few people could innovate, then Apple came along and revolutionized this model by creating a device with open interfaces where it was, you know, a lot of people could innovate and the cost of application innovation were really high. So in line with my predictions, moving from a world where cost of application are high, only a few people can innovate to a world where cost of application are low, a lot of people can innovate and a lot of different people we have, Stanford students who write applications for the iPhone. I know lots of people who write applications for the iPhone in their free time. It's not a surprise that we get a lot more application innovation. So it's not a surprise that we get a lot more application innovation. And of course, there is one factor that differs which is the control. But there we now have the interesting experiment that we have the iPhone and Android where both have low cost of innovation, both allow a wide variety of innovators to innovate, but Apple controls the system and has final say over who gets on the platform. Google doesn't. And so to me, this is the point where we see how these predictions play out. And my initial take is that, you know, originally Apple basically wrote on this idea that on the fact that there were so many more people using the iPhone than people using Android. But I think we're starting to see this shift where lots of innovators in Silicon Valley and me tell me, you know, I always start developing an application for Android because once I'm done, I put it out there, I start realizing revenue. And then I decide whether I also want to develop for the iPhone because if I develop for the iPhone, first I need to wait and I don't have any income while I wait. And then I might even not be successful in getting on the iPhone. And I know some people who have made the decision not to program for the iPhone anymore. So I think we have actually seen two shifts and both are in line with the theory. The first shift, lower the cost of innovation, increase the amount of innovator, great explosion innovation. And now this idea control versus no control. And I think we are seeing that the platform with no control is starting to be more successful. And as to the wireless network, I totally think the same rules should apply to wireless. You know, it's as important. The same analysis applies. And to the extent there are some technical differences between wireless and wireless networks, we can take account of them in the network management exception. So, you know, if we were talking longer, we would talk about the fact that every network neutrality rule has a network management exception. And in my proposal, this exception would say, you know, if you need to manage your network as application agnostic as possible, and the as possible is the safety valve, and it may be that in some form of, in some wireless networks, it's not possible to do things in an application agnostic way, in others it may be, and so that's the point where we take account of these differences. Okay. Thank you for your talk. I just have two objections. The first is that the argument of the importance of net neutrality might be a little deceptive because the network is not the only thing that matters. If I were to create a competition for YouTube, I would need to create this huge data warehouse, and if the network is neutral or not, that doesn't change my problem at all, so I still need a lot of money. Similarly, if I were to create a competition for Facebook, network effects and the number of people already in Facebook would be a huge impediment to my success. So just having the network up and doesn't solve the problem. The second objection is that why is innovation the only value we need to uphold in a sense that might there be some sort of innovation that's not low cost and that might benefit from protectivism? And in kind of a similar fashion, what happens when innovation is actually hurtful to the rich incumbents like the news channels and people have discussed a lot about newspapers, not being able to survive, what happens when the innovation that is low cost hurts something that is socially valuable? Thank you very much. Okay, so your first point about an open network not solving everything and I thought there were actually two points in that part of your question, which was there might be additional things you need, but also I thought you were questioning how effective it was if the network enabled you to innovate at low cost because there might be other things that might create barriers to entry which actually are associated with higher costs. Yeah, so I agree. If you are a rich company, you may have a bigger marketing budget and as a new entrant, it may be more difficult for you to reach people. So all over the world, or if you come from rich parents, you may have a better education that may enable you to program better. So there are all sorts of inequalities in the world which may create differentials, but to me, the network itself is special because it's basically the fundamental point of entry for innovators into the network. And even if there is a new class of applications that comes up, they aren't yet afflicted with network effects. So Facebook, when Mark Zuckerberg founded Facebook, he was a grad student and in my book I have the exact number of how much he wasn't a grad student, he was an undergrad at Harvard. And I have the numbers of how much it cost him to actually run Facebook and it was insanely low, the cost of a server running per month. And there were some existing social networks, I mean MySpace and Friendster. So it's not as if he wasn't up against network effects but still he was very successful and he only got venture capital. And again, the exact numbers on the book when he already had a very significant number of users. And so I mean, I was about to say, at least in new classes of applications you don't face these problems of network effects. But the Facebook example shows that even in areas where there are existing applications you may as the new entrants be able to enter. And so I do think this is very important, a very important feature of the internet to preserve. Then innovation may be disruptive, yes, that's a feature of innovation. And it may be disruptive to industries that we care about. But I don't think that's a reason for messing with this fundamental infrastructure that allows innovation in all sorts of different areas. I think that's a reason to say, okay, we may have a problem with newspapers here. And so let's think about how we can make sure that newspapers continue to exist in one form or another. Or maybe we don't need newspapers anymore. You know, if you read your high stuff then maybe we can have other arrangements that realize the same values. But I think you can separate that problem from the infrastructure. And as a general matter, I think that's exactly one of the good things that the way the internet is structured, it allows new entrants to disrupt incumbents. And I would never want to help the incumbents protect their business models by establishing barrier towards innovation in the network. Hi, you mentioned that you want to charge users but not providers in the network. Obviously in a peer-to-peer setting or similar settings, the distinction between a user and a provider is blurred. So how do you propose that we define users and providers? Thank you. So you are a user if you are the network provider's internet service customer, access customer. So all the access network customers are the users, no matter for what they use the network, whether it's to send email or participate in peer-to-peer making content available, that's not the difference. The difference is your commercial relationship in the form of a contract for internet access network services. So if YouTube was a customer of Comcast and it wanted better service, it could buy quality of service from Comcast but only on Comcast's network. But if YouTube had let's say level three as their provider or Google's network, then Comcast couldn't say you're not attached to my network but in order to have your data packets travel to my internet access customers, you need to pay. So that's the distinction. So if you invited your grandmother, get on the internet, you do your medical monitoring, you can be connected to the world and you're mobile. She can't come to this lecture. So how do you fit the inability to communicate at all into the idea of non-discrimination? I wouldn't think of that as a problem of non-discrimination. I would think of that as a separate problem that's linked to the question of how do we get infrastructure and who can get access to infrastructure. And right now we live in a world where you get access to this infrastructure through contracts or affiliation with network owners. And if you're a Harvard student, you probably are able to get to the internet in this room. And I'm happy to think about alternative models of doing this but I really don't think it has anything to do with network neutrality. Okay, so I'd agree with your whole premise that innovation is the most important thing because I think as an American it's the only thing we have going for our country anymore. Certainly not our value of education or our health or anything. But the one argument I've heard against your point of view which is not a fundamentalist free market argument is contained in your anecdote about your grandmother which is that not every consumer necessarily wants to pay what something is worth. And sometimes that has beneficial effects. So you could have subsidies like classified ads, subsidize the newspapers or people will pay a lot of money for cable TV content but they wouldn't necessarily pay for that connectivity to their house. So how do you preserve that effect or allow business models that might charge for some service or some application or some content to subsidize some of the capital investment that might be required to continue to upgrade the quality of the network? So I just want to reframe your point or I restate it and so I take you to say, well isn't internet access a little bit like newspapers or could it be like newspapers? And we pay less for the newspaper because advertisers pay money to the newspaper and that allows the newspaper to subsidize prices to readers of the newspaper. And that's not how the internet currently works but you might think that might be a good idea. And so it would run against the YouTube story and this idea that network providers aren't allowed to charge application and content providers. So there is an immediate answer to this question which applies to the US, which is the theory of two-sided markets which determines what happens in this space where you have an intermediary, the internet service provider and then two sites in the newspaper case, the newspaper readers and the advertisers in our case, the users and the application providers who want to get together and the platform intermediary does that and then decides how to charge prices to each side of the market. So this theory predicts that if there is, if one of the sites of the market is single home that means only attached to one site, to one specific platform, then the network provider has an incentive to charge monopoly prices to the multi-homing site, to the multi-homing site. And in our case, single home, most users have one internet access provider so they are single home and a company like Google wants to reach customers all over the world so they are attached to lots of different access providers they are multi-homed. And this theory predicts, as I said that this gives the network provider the ability to charge monopoly prices to Google. So that's one of the things that's bad within this arrangement. But there is another interesting aspect of this model that speaks directly to your question which says under which conditions does the network provider then have an incentive or the platform intermediary then have an incentive to transfer some of the profits that it makes on the multi-homing site to the single-homing site. And the theory says, well, if there is competition, well, look around yourself. I don't know where exactly you live but in most parts of the US we don't have competition among access providers. And that suggests that at least if these models are right, the network providers wouldn't actually use the additional profits to subsidize prices for internet access but they would use it to give higher dividends to their shareholders or other things that they might want to do with the profits. Beyond that, let's say we were in Europe where there is more competition. I think you then face a real trade-off where you need to ask yourself, well, if allowing network providers to charge application and content providers may bring down the prices of internet access or if there was a variant of this argument which says the network providers would use the additional money to deploy more broadband networks, wouldn't that be great? And so basically you get to a trade-off where you say it's application innovation on the one hand and then more and better broadband networks or lower prices for internet access on the other side and that's a difficult trade-off because it's very clear, without access to the internet you can't really benefit from all the advantages. And so I think about this trade-off in the following way. I think, well, what would happen if we forced the network to be neutral and not charge application and content providers? Would we be able to solve this problem for network deployment in different ways? Yes, we would because the problem then is we need more money. Our network providers need more money and so we need to find money which of course isn't totally trivial but we could do it and give it to them through subsidies, text breaks, subsidies to internet users, universal service fund is one example. So a solvable problem. Now let's think about the other alternative. Let's imagine we allow network providers to charge access charges. Can we then solve the problem for application innovation and free speech and all the other good things we care about in different ways? No, we can't because we break the mechanism that created application innovation at its core. Suddenly people can't contribute to political discussions at low cost anymore because content providers with deep pockets have an advantage and that tells me it doesn't make sense to give up what's valuable about the internet to solve a problem that we could solve otherwise. And Tim Vue has this nice site that I really liked where he said it's like selling the picture to get a better frame and I really don't think that makes sense. I actually have two questions but how do you guard against, if you allow discrimination based on quality of service, how do you guard against the creation of an artificial distinction? For example, say a network provider has one quality of service, this is called managed isochronist delivery and has another managed isochronist delivery. And another one, wait a second. Then the second one is called unmanaged isochronist delivery. So isochronist means that you're providing a certain bandwidth over a given period of time and you can use that for delivering video. So managed delivery of video and unmanaged delivery of video. But you explicitly use a quality of service term like isochronist delivery to obfuscate the fact that what you're doing is discriminating against one form of video and not against another form of video. And you can take that more explicitly and say, a cable channel is a isochronist managed form of delivery and YouTube over an internet link going over the same infrastructure is an unmanaged isochronist delivery of data. Can you just explain to me why it would, this type of thing would favor certain types of video over others, which one would be favored? No, you'd use that as a point of discrimination. But the problem is- You're using quality service. Yeah, I get that. You're using quality service acronym. But you don't get my proposal I think because in my world you couldn't use that as something to discriminate on because the network provider would just offer these two types of service. And then I could choose which one I want to use for my video. And so at least the network provider couldn't push me in a certain direction. You might say, well, but usually it will probably cost you something to use this higher quality of service. And that creates a different set of problems that I think you weren't thinking about that, but that you should be thinking about, which is the moment we allow the network provider to charge for better service. It creates an incentive to the network provider to artificially degrade the quality of the baseline service. And the best way to see this is to think of airplanes. So airplanes or airlines have an incentive to make economy class sufficiently unattractive in order to motivate rich people to pay for business class. And that's not just an anecdote, there is actually sound economic theory behind this, the theory of price discrimination. And so that's a very real problem that we would face. And so that the network provider would have very crappy baseline service and then everybody needs to pay for video and or whatever else they need the service for. And I agree it's a real problem and many people use it to argue we shouldn't have quality of service at all. The Europeans have a solution that I find interesting which is to say, yes, this problem exists. So we give the regulator the ability or the obligation to monitor the quality of the baseline service and impose minimum standards for service if the quality drops below acceptable levels. And this of course has a whole bunch of problems which are, how do you define below acceptable levels and what exactly, how can you do that in good ways? And I don't say these problems don't exist. The Europeans think that just the threat of doing something will be sufficient to discipline network providers and right now I think that I really wanna allow the internet to evolve and so I would go in favor of this sort of monitoring and potential intervention solution over saying no quality of service at all because in the end there might be some applications that really require strict guarantees and I wouldn't want to lose out on them. Let's look at the last question. Thanks very much.