Sunday, July 27, 2008

A Reply to Nachbar

All right, I'm in final stages of thesis editing and I thought I'd post some excerpts (excuse the still-rough footnotes). In today's excerpt, I discuss Thomas Nachbar's recent article The Public Network. I had initially read Nachbar's earlier version of this paper, which he presented at TPRC 2006. This version was entirely a historical accounting of open access mandates in communication law (or non-discrimination, depending on your personal preference). The paper was absolutely fundamental in my understanding of communications law. In this latest version, he adds substantial sections in which he analyzes present debates in net neutrality and spectrum auctions in light of his historical account. I disagree strongly with his conclusions.

Thomas Nachbar has argued that the ideal non-discrimination rule would prevent user-based discrimination but allow carriers to discriminate based on use. Under this regime, providers would be able to choose which services they support (and how they prioritize or discriminate among them) but they would be required to offer the same deal to everyone. Google could not pay for faster delivery than Yahoo. He reasons that user discrimination is easier to define than use discrimination, and less prone to regulatory abuse. He envisions this user-based neutrality as enforced by “standards” and not law or formal rules.[1] Furthermore, he claims that mandating uniform treatment of all packets would discourage applications that require prioritization or quality of service guarantees, making it a type of discrimination itself. To be sure, networks that treat all traffic uniformly make it more difficult to use certain applications. However, Nachbar’s core criticism appears to be not that someone will be choosing how to prioritize, but rather that in some neutrality regimes the government would be choosing. The best entity to choose, on his account, is the last-mile provider.

I disagree. Both use and user non-discrimination should be policy goals. It makes sound economic sense, it is consistent with historical non-discrimination precedent, and supports internet ethos of diverse uses and abundance of peers. Historically, use and user they were closely linked, and non-discrimination in one area could ensure non-discrimination in the other. For example, the Computer II rules mandated only that carriers not discriminate based on the phone number called. However, because of the simplistic circuit-switched technology (and the Carterphone right to attach devices), the rules ensured that use-based discrimination would not occur. Today, user-based discrimination protects only against a subset of harms, which in any event might already be addressable under antitrust doctrine.[2] It does not ensure that carriers support applications that they do not think will be profitable, or that compete with their non-internet offerings,[3] or that have not yet been invented. The problem is that surrendering use-based discrimination to last-mile providers would subject the general-purpose infrastructure in the interest carrier-profit-oriented incentives. In fact, it discriminates against users with business models or non-commercial modes of production that rely on technology uses not approved by the carrier.[4] The technology of the internet presents us with a choice we have not had to make historically because user-based neutrality has always implied use-based neutrality. Nachbar is prepared to give up on use neutrality, while I am not.

One way to maintain use-based non-discrimination by carriers would be to place prioritization control in the hands of the users. Most content/application providers have the opportunity to exercise this control by going to any number of competitive backbone providers. Different backbone providers ensure different levels of quality-of-service guarantees for common metrics like latency, throughput, and jitter (at least, up to the edge of their networks). End users, who are accessing this content or these applications or are connecting with each other in peer-to-peer fashion, do not have the ability to choose different prioritization via competitive providers or by specifying preferences to their provider. Indeed, even across-the-board neutrality may disfavor particular applications users wish to use, although this may be more appropriate and efficient than the last-mile provider’s blanket imposition of prioritization. A better solution would allow end users to easily control the prioritization of their own traffic, within the tier of service that they have purchased from their provider. Such a solution might implement a more sophisticated “Type of Service” style component into some layer of the network protocol, after being defined via a standards group such as the IETF.[5] This approach recognizes that different users have different usage needs, and places the control in their hands. It refuses to foreclose on new uses simply because the network owner did not think of them first, and catalyzes innovation at the “edges.” It is not true to absolute neutrality, but it is true to fundamental principles of non-discrimination and the internet ethos. Such an approach is unlikely to garner initial favor with carriers because it preserves user control, because it nevertheless resolves their “congestion” justification, and because it would take more technical and cooperational work than blunt discrimination. The appropriate policy path to this outcome might involve a use-neutrality mandate on last-mile providers with an exception for user-specified, standards-defined prioritization.

[1] It is unclear what these “standards” might be, other than the existing standards within the internet protocol, which have clearly been ignored in cases such as the recent Comcast/BitTorrent back-and-forth. As such, I am not sure what real force they would bring to bear on the situation aside from the unsustainable ad hoc complaint adjudication that the Commission is currently undertaking.
[2] I am skeptical and discuss this in detail earlier in the thesis.
[3] Although Nachbar seems to think so (eg telephony and video)
[4] eg, p2p
[5] Reed at the Harvard hearing. “There were a wide range of actual standards that would allow Comcast to manage and prioritize traffic, including diffserv, ECN, RED...” One might add to this list the RSVP protocol (RFC 2205) and other methods that use flow-based prioritization (such as the method described in J. L. Adams, L. G. Roberts, A. Ijsselmuiden, Changing the Internet to support real-time content supply from a large fraction of broadband residential users, BT Technology Journal, v.23 n.2, p.217-231, April 2005). Some of these tools can be used by network operators to choose their own discriminatory practices, or they might be implemented in such a way as enable user-based control. Early internet engineer David Clark recently remarked (video recording available at with quote at 4:24:45)

I don't like the idea of the ISP assigning quality of service to an application. If there is going to be any discrimination in terms of quality of service that's associated with some packets rather than others, I would prefer that the bits which select those packets for enhanced service be set by the user. The user could say 'this telephone call is really important. I want this telephone call to go through.' Imagine that in any given month, ten percent of your traffic could be high priority. You could say, 'this is it, I want it here.' It could be my choice as to whether that's a phone call or a game, or I'm trying to get a bid into eBay or whatever I'm trying to do. I would like the user to be able to assign those priorities. If you look at the way that internet telephony is done today, those bits are set by the phone device. It's not set by the ISP. It's the phone device that says, 'this is a phone call and therefore I will set these bits,' and if the ISP chooses to honor these bits then these packets will go through better. That's something that could be superimposed on top of the basic idea of usage quotas.


No comments: