End-to-end and the Price Mechanism

by Tim Lee on May 1, 2008 · Comments

One of the more positive consequences of the whole Sydnor/Lessig debate is that it’s enticed the always-interesting David Friedman to weigh in on tech policy issues, giving me the opportunity to quote him in his entirety:

I’ve been trying for years to persuade Larry to admit to libertarian leanings. I’m not sure from comments here whether he actually did it while I wasn’t looking or is merely being accused of it. My interpretation of his attitude, long ago, was libertarian instincts hindered by a leftist self-image.

Consider his basic argument, a book or two back, for treating the net as a commons–by which he actually meant a commodity, since he wasn’t proposing zero cost access. It was that if the people in the middle, the ones transmitting the bits, got a veto over what sorts of bits they transmitted, that would make innovation very hard, since there would be too many people whose agreement you needed before doing anything.

I think he acknowledged–certainly he knew–that the counter argument was that what was being transmitted varied in ways that were relevant to the cost of transmitting it–burst vs steady stream, material where very low lag was important (real time games, distance surgery) vs material where it wasn’t (downloading), etc. So requiring the same cost for everything, or even specifying the cost structure, meant that some people were free to impose external costs on others without their consent. His conclusion depended on the judgment that the inefficiencies due to permitting that were less important than the inefficiencies due to the high transaction costs of innovation with the alternative system.

What didn’t seem to occur to him was that he had just sketched the argument against zoning. There too, the individual’s decisions–what sort of house to build, whether to use his land for residence or commerce, and the like–can impose external costs on others. There too, requiring the permission of those others, whether directly or via variances in zoning, makes innovation hard. The same argument Larry was making for the net as a commodity, applied to land use, is an argument for strong individual property rights and against land use control. Once you take seriously the point that forcing people to take account of all effects of their acts on others means nobody ever gets to do anything, you undercut a lot of the arguments for a wide variety of government interventions.

Quite so. I’ve said before that network neutrality (the technical principle, not the proposed legal regime) is the division of labor. The end-to-end principle allows decentralized decision-making on the Internet in precisely the same way that the price mechanism allows decentralized decision-making in the broader economy: by giving people a simple, predictable interface to the rest of the world that isn’t dependent on the whim of any central decision-maker.

ISPs that try to implement discriminatory network policies create the same kinds of problems as government officials that enact regulations: they often cause unintended consequences (like blocking Lotus Notes) and they cause people to waste resources evading the restrictions (as with BitTorrent header encryption).

Now, I should hasten to add that as I’ve written before, the fact that neutral networks have good properties doesn’t mean that mandating them is good public policy. Because of course a network neutrality rule would itself have unintended consequences and lead people (in this case ISPs) to waste resources trying to evade the rules. But if we’re talking about network design, rather than government regulation, it seems to me that libertarians ought to look favorably on decentralized networks mediated by the end-to-end principle for all the same reasons we look favorably on decentralized economies mediated by the price mechanism.

Comments Posted in: Uncategorized

  • See Webex Mediatone network, a global IP network that provides quality of service guarantees. This is what we call an "overlay network" that intersects with the public Internet at several places, but only carries traffic for subscribers.

    Regarding your second question, I'm actually in favor of a system where users decide which traffic gets which treatment, but there are some interesting questions to be resolved about the generic fast and slow lanes. I don't see any need for carriers to degrade Vonage in order to make their own voice products work better; the pirates are the ones doing the degrading already, so the carrier can either stop that or allow it to continue. The typical NN regulation scenario would prevent them from stopping it, such that the pirates would freely degrade Vonage and Skype more or less at will. This is what happens when lawyers try to design networks.
  • Ryan Radia
    Richard, you make an interesting point about differential fees for traffic depending on the importance of time sensitivity.

    Is there currently a pricing mechanism being used that allows backbone providers to charge based on both bandwidth and latency? Is there some sort of global internet expressway where the charge per byte is higher but the time to traverse the network is lower versus the regular network? Do Tier 2 carriers pay backbone providers differential rates for traffic depending on how long each packet takes to get from one end to another?

    Also, how do you view the distinction between application neutrality and endpoint neutrality? Many NN advocates are fine with ISPs prioritizing VoIP over web, but they aren't fine with ISPs doing something like degrading Vonage to push customers towards in-house offerings.
  • Matt: quite so. I think Lessig wants ISPs to act as neutral referees that provide a solid foundation for the application market but don't themselves compete in the application market. You can once again draw an analogy to the free market, in which the state is a neutral referee enforcing the rules of property and contract without trying to favor some firms over others.
  • Matt S
    Thanks Tim, I get you. I think your prediction is probably right, that open will win -- not because it's "good and right", but because the market will not likely sustain other models. (And, always appreciated when you reiterate that this is a desired outcome and not a regulatory suggestion.)

    It sounds like Lessig's ideal is that "the Internet" would be a frictionless space in which the market resides, but not a participant in the market itself? Is that your reading?
  • Now we're into the shifting sands of the Internet regulation debate, where the terms "end-to-end," "neutrality," and "openness" compete with each other for air time. They're not interchangeable.

    Lessig started this discussion by trying to draw policy implications from E2E. Tim Wu came along and observed that E2E has a built-in bias toward a particular class of application (non-time-senstive file transfer applications, specifically,) and the pro-regulatory crowd responded by changing their goal to "openness."

    If we want networks that support user experimentation, which is what "openness" implies, we need to have enough control inside the network to prevent experimental applications from interfering with each other and with mainstream network usage. The Internet has never had this, which is why its real history is a series of crises.

    A network architecture that fully supports openness needs to support an interaction between the network and the application where the app tells the network what it needs and the network tells the app what it can supply. The two agents then need to negotiate with each other for a service level and price that satisfies both, and it's only in this interface that it makes sense to introduce market talk.

    In an ideal world, a user would not be bound to his one ISP for some sort of default delivery service across the public network. In fact, he would be able to choose network delivery services application-by-application that would serve his requirements for price and performance. On the PSTN, we can choose from a number of long-distance carriers call-by-call, so we can use one service to call New York and another to call New Delhi if that's what we want.

    On the Internet, we do this through overlay networks that can provide low-delay delivery service for a relatively high per-packet fee when we need it, and high volume but high delay services for a lower per-packet fee.

    There is a rough analogy between markets and E2E because Lessig has worked very hard to create the appearance of similarity. But it's a false and misleading analogy for a number of reasons.
  • Matt, thanks for commenting.

    My claim is certainly not that libertarian ideology requires support for open networks. Clearly it doesn't. Rather, my point is that Lessig's basic argument for end-to-end is closely analogous to Hayek's argument for free markets, namely that decentralizing decision-making will tend to lead to more efficient outcomes because individuals possess local knowledge that cannot easily be aggregated and transmitted to a central decision maker. And while a network owner obviously isn't a central planner in a political economy sense, it clearly is a centralized authority within the context of the network. That is, in a proprietary network, users will often only be able to use those applications that the network owner has specifically authorized. To the extent that the end-to-end principle decentralizes decision makers away from network owners and toward end users and application developers, I think we should see beneficial results.

    Now, of course, this isn't a perfect analogy and libertarians can quibble with the details. It's conceivable that, as Richard suggests, there are applications that will only perform well if they receive certain kinds of special treatment from network owners. And of course, if we're talking about public policy, there's a strong libertarian presumption in favor of allowing network owners to experiment with different business models and let the market decide which one is the best.

    However, the fact that we don't want the FCC telling network owners how to run their networks doesn't mean we can't have opinions about which networking architectures are likely to win in the marketplace. My guess is that the open networks will win, for the reasons I laid out above. Whether you want to call these "libertarian" reasons or just reasons that are coincidentally parallel to common libertarian arguments is little more than a semantic issue. My point is simply that there are striking parallels between the arguments for free markets and the arguments for end-to-end, and that these parallels are not entirely coincidental.
  • Matt S
    I don't see how libertarians should have an instinct for or against neutrality as a principle. Seems to me a libertarian would want the market outcome. There may be market outcomes that are beneficial to some consumers, yet decidedly non-neutral. Some people may be perfectly happy with walled gardens.

    Seems to me the essentially libertarian idea is that markets invent things we can't predict, and that they must be allowed to do so.

    How do we know that an end-to-end network is and always will be superior, for all consumers? Shouldn't all possibilities be tested in the market, and shouldn't libertarians desire a plurality of outcomes?

    Are where does this centralized authority idea come from? ISPs are not a centralized authority, they are private businesses, pursuing their own interests, that choose to interoperate.
  • You make the same error that Lessig makes. The end-to-end model is one way to design a network, but it's not neutral. End-to-end is a bias in favor of applications with loose time requirements, and that's not the universal pool of applications.

    In fact, the only way to design a truly neutral network is to use end-to-end for some things and hop-to-hop for others, depending on their needs, and other types of mechanism for other things. Diversity, in other words, is more "neutral" than the Internet Protocol monoculture proposed by The Big Demagogue.

    So please don't jump from E2E to neutrality; they're not really the same.
blog comments powered by Disqus

Previous post:

Next post: