Live blogging Barbara van Schewick’s talk at Maxwell Dworkin here at Harvard. (That’s the building from which Mark Zuckerberg’s movie character stumbles through the snow in his jammies. Filmed elsewhere, by the way.)
All the text is what Barbara says, or as close as I can make it. My remarks are in parentheses. The talk should show up at the MediaBerkman site soon. When it does, go there for the verbatim version.
(In the early commercial Net, circa 1995 forward), the innovator doesn’t need to ask permission from the network provider to innovate on the network. Many different people can innovate. Individuals at the network’s ends are free to choose and to use. Obligation to produce a profit in the future isn’t required to cover development costs, because those costs are often cheap.
Innovators decide, users decide, low costs of innovation let a large and diverse group can participate.
The network is application-blind. That’s a virtue of end-to-end. (Sources Reed, Saltzer and End-to-End Arguments in System Design.)
Today the network operators are in a position to control execution of programs. “Imagine you have this great idea for a video application… that means you never have to go back to cable again. You know you have a fair chance at the marketplace…” In the old system. Not the current one. Now the network provider can stand in the way. They say they need to manage bandwidth, or whatever. Investors don’t invest in apps or innovators that threaten the carriers directly.
Let’s say Google ran the network when YouTube came along. Would YouTube win this time, like it did the first time? (Disregard the fact that Google bought YouTube. What matters is that YouTube was free to compete then in ways it probably would not now—so she suggests.)
In the early Net (1995+), many innovators decided, and users decided. There was little uncertainty about the supportive nature of the Internet.
User uncertainty or user heterogeneity? More and better innovation that better meets user needs. More ideas realized. (That’s her slide.)
With fewer or less diverse innovators, fewer ideas are realized.
Her book concentrates on innovators with little or no outside funding. (Like, ProjectVRM? It qualifies.)
One might ask, do we need low cost innovators now that there are so many billionaires and giants like Google and Yahoo? Yes. The potential of the early Web was realized by Netscape, not Microsoft. By Amazon, not by Barnes & Noble.
Established companies have different concerns and motivations than new innovators. Do we prefer innovation from large self-protecting paranoid companies or small aggressive upstarts?
Users decide vs. Network providers decide. That’s the choice. (The latter like to choose for us. They did it with telephony and they did it with cable TV.) In Europe some network providers prohibit Skype because it competes with their own services. Do we want them to pick winners and losers? (That’s what they want to do. Mostly they don’t want to be losers.)
Users’s interests: Innovators decide. Users decide Network can’t control Low costs of innovation, very large and diverse group of innovators. (Her slides are speaker’s notes, really.)
Network providers’ interests: They are not interested in customer or user innovation. In fact they oppose it. They change infrastructure to protect their interests. There is a gap between their private and public interests: what economists call a Market failure.
Do we need to regulate network providers? That’s what Network neutrality is about. But the high cost of regulation is a difficult question. Not saying we need to preserve the Net’s original architecture. We do need to protect the Net’s ability to support innovation.
Let’s pull apart network neutrality and quality of service (which the carriers say they care most about).
Best effort is part of the original design. Didn’t treat packets differently. Doing that is what we call Quality of Service (QoS).
Question: How to define discrimination? We need to ask questions. Such as, do we need a rule against blocking? Such as against Skype. One defining factor in all NN proposals is opposition to blocking. If Comcast slows down YouTube or something else from Google to favor it’s own video services (e.g. Xfinity), that’s discrimination.
Option 1: allow all discrimination…. or no rule against discrimination. That’s what the carriers want. Think of all the good things you could get in the future that you can’t now if we allow discrimination, they say. (Their promise is a smooth move of cable TV to the Net, basically.)
Option 2: ban all discrimination … or treat every packet the same. This is what Susan Crawford and others argue for. Many engineers say “just increase capacity,'” in suipport of that. But that’s not the best solution either. It’s not the job of regulators to make technical decisions about the future.
All or nothing doesn’t work. Nether allow all discrimination nor Ban all discrimination.
Application blindness is the answer.
Ban discrimination based on applicaitons. Ban discrimination based on applications or classes of aplicaitons.
Fancast vs. Hulu. YouTube vs. Hulu. Allow discrimination based on class of aplication… or like treatment. Treat internet telephony vs. email differently. But don’t favor Skype over Vonage. (This is hard to describe here. Forgive.)
Problem 1. Distorting competion. Capturing some value from gaming, for example, by favoring it as a class. Give it no-delay service while not doing that for VoIP. But both are affected by delays. In the Canadian network management proceding, we found that P2P is slowed down either all the time or during congestion time. That allows real-time to work well. But then real-time video came along. What class do they say that belongs to? We don’t really know what the Canadian carriers did, but we do visit the question of what they should do if they discriminate by class. Thus…
Probem 2. High cost of regulation. (Self-explanatory, so it saves me the effort to transcribe.)
Problem 3. User choice. Support from the network. The moment you require support from the network (as a user or app provider), you throttle innovation.
Constraints on Network Evolution allows quality of service: 1) Dfferent classes of service offered on a non-discriminatory basis; 2) Users able to choose wheter and when to use which class of service; 3) Net provider only allowed to charge its own Internet service sustomers for use of different classes of service*. So network providers don’t destroy competition any more. Users get to choose which quality of service to use. And the network provider doesn’t need to provide QoS except in a general way. They’re out of the market equation.
(Bob Frankston is across the aisle from me, and I can see the word balloon over his head: “Why constrain thinking with ‘services’ at all? Why not just start with connectivity? Services keeps us in the telecom bottle.”)
Constraints on network evolution. Cost of regulation.
MY SOLUTION: (not on screen long enough.. there was more on the slide)
Preserve factors that have fostered application innovation ≠ Preserve original archictecture of the internet.
Final question to talk about. Why care about application innovation?
Have you ever tried to explain to your partner’s grandmother why she should use the Internet? You don’t argue about sending packets back and forth. You talk about grandchildren pictures, and being able to talk for free. That comes from innovation at the ends, not the carriers.
We need to protect the sources of innovation.
Yochai: What do you do with Apple iPhone? Tremendous user adoption being driven precisely by a platform that reverses many of your assumptions smack in the middle othe most controversial boundary, regarding wireless. (Not verbatim, but as close as I could get.)
Barbara: People say, “Look, I’ve got a closed device supporting lots of innovation.” No, you need to think about this differently. Apple created a device with open interfaces that supported lots of innovation. So it moved us from a world where few could innovate and it was costly, to a world where many could and it was cheap. Proves my point. Now we have an experiment with iPhone vs. Android. Apple controls, Google doesn’t. Now we get to see how this plays out. We’re starting to see where lots of innovators are moving to Android as well. More are starting with the Android, experimenting and then moving to the iPhone. The cost of starting on the Android is less. So we have two shifts. I think we will se the platform with no control being more successful.
Every network neutrality proposal has a network management exception. Mine doesn’t.
Q from the audience; Some apps still need a lot of money, whether or not the network is neutral. Building a big data warehouse isn’t cheap. And why is innovation all that matters? What happens when it is actually hurtful to rich incumbents such as news channels?
Barbara: I agree. If you’re a rich company, your costs of entry are lower. Kids with rich parents have advantages too. To me the network itself is special because it is the fundamental point of entry into the marketplace. We want the impediments to be as low as possible. The cost of starting Facebook for Mark Zuckerberg was actually rather low. He scaled after getting VC money, but he got a significant number of users first, without a lot of costs. I do think this is very important. Innovation is often disruptive, sure. But that’s not a reason for messing with this fundamental infrastructure. If newspapers have a problem with the Net, fix the papers. Separate that problem from the infrastructure itself. As a general matter, one of the good things about the Net’s infrastructure is that it allows disruption.
Q: What about companies as users? (Can’t summarize the answer.)
Bob Frankston: If your grandmother is on a phone… (couldn’t get what Bob said or make sense of Barbara’s response… sorry).
Q: (What about subsidies? I think.) The theory of two-sided markets. With papers, subscibers and advertisers. With the Net, users and app providers. If you’re attached to one platform, the providers are likely to attach to one side. (I think that’s what she’s saying.) This gives the provider a way to monopolize. In Europe, where there is more competition, there are more trade-offs. I think what would happen if we forced the net to be neutral, would we solve the problem by charging a different way. Subsidies, tax breaks. Perhaps a solvable problem. Let’s say we allow the carriers to charge extra (for premium use?). We break the system at its core. It doesn’t make sense to give up the value of the Internet to solve a problem that can be solved a different way.
Q: A question about managed vs. unmanaged isochronous delivery. We should be thinking about what happens when the carriers start charging for better service. (But they already do, with service tiers, and business-grade service (with assigned IP addresses, unblocked ports, etc.). The Europeans give the regulators the ability to monitor quality and impose minimum standards. This has a whole bunch of problems What really are acceptable levels? for example. The Europeans think this is sufficient to discipline providers. Well, in the end there might be some apps that require strict guarantees.
Okay, it’s later now. Looking back over this, I have to say I’m not sure it was a great idea to live-blog it. There are others who are better at it. Within the Berkman fold, David Weinberger is one, and Ethan Zuckerman is another. Neither were in the room, so I thought I’d give it a try. Again, visit MediaBerkman for the actual talk. Or just go get her book, Internet Architecture and Innovation. I got one, and will start reading it shortly.
The picture above, by the way, is one of a set I shot at the talk.