Network Neutrality, Broadband Discrimination

Columbia University law professor Tim Wu coined the term “net neutrality” in a paper he published in the Journal of Telecommunications and High Technology Law. The paper is an interesting read because it’s sharply opposed to the regulations adopted by the House Judiciary Committee this week, so I’d encourage anyone who wants to have a neutral Internet to go read it. Some of Wu’s more interesting observations follow.

Proponents of open access have generally overlooked the fact that, to the extent an open access rule inhibits vertical relationships, it can help maintain the Internet’s greatest deviation from network neutrality. That deviation is favoritism of data applications, as a class, over latency-sensitive applications involving voice or video. There is also reason to believe that open access alone can be an insufficient remedy for many of the likely instances of network discrimination…

I believe there are several reasons to question the fit between open-access remedies and network neutrality. First, the concept of network neutrality is not as simple as some IP partisans have suggested. Neutrality, as a concept, is finicky, and depends entirely on what set of subjects you choose to be neutral among. A policy that appears neutral in a certain time period, like “all men may vote”, may lose its neutrality in a later time period, when the range of subjects is enlarged.

This problem afflicts the network neutrality embodied in the IP protocols. As the universe of applications has grown, the original conception of IP neutrality has dated: for IP was only neutral among data applications. Internet networks tend to favor, as a class, applications insensitive to latency (delay) or jitter (signal distortion). Consider that it doesn’t matter whether an email arrives now or a few milliseconds later. But it certainly matters for applications that want to carry voice or video. In a universe of applications, that includes both latency-sensitive and insensitive applications, it is difficult to regard the IP suite as truly neutral as among all applications.

This point is closely linked to questions of structural separation. The technical reason IP favors data applications is that it lacks any universal mechanism to offer a quality of service (QoS) guarantee. It doesn’t insist that data arrive at any time or place. Instead, IP generally adopts a “best-effort” approach: it says, deliver the packets as fast as you can, which over a typical end-to-end connection may range from a basic 56K connection at the ends, to the precisely timed gigabits of bandwidth available on backbone SONET links. IP doesn’t care: it runs over everything. But as a consequence, it implicitly disfavors applications that do care.

Network design is an exercise in tradeoffs, and IP’s designers would point out that the approach of avoiding QoS had important advantages. Primarily, it helped IP be “downwardly” neutral as to the underlying physical media. But this requires us to be more circumspect in our discussions of network neutrality. IP’s neutrality is actually a tradeoff between upward (application) and downward (connection) neutrality. If it is upward, or application neutrality that consumers care about, principles of downward neutrality may be a necessary sacrifice.

This returns us to the question of structural separation. We have a public network that is indeed a great creative commons for data applications, but it is less so for any application that requires a minimum quality of service. True application neutrality may, in fact, sometimes require a close vertical relationship between a broadband operator and Internet service provider. The reason is that the operator is ultimately the gatekeeper of quality of service for a given user, because only the broadband operator is in a position to offer service guarantees that extend to the end-user’s computer (or network). Delivering the full possible range of applications either requires an impracticable upgrade of the entire network, or some tolerance of close vertical relationships.

This point indicts a strict open-access requirement. To the extent open access regulation prevents broadband operators from architectural cooperation with ISPs for the purpose of providing QoS dependent applications, it could hurt the cause of network neutrality. By threatening the vertical relationship required for certain application types, it could maintain IP’s discrimination in favor of data applications. More broadly, this argument shows that the concept of network neutrality cannot be taken as counsel against all vertical integration.

So which side of the debate is Wu on? He argued before Congress that the Telcos are acting like gangsters, and Congress addressed his concerns by making the Internet less neutral than it is today.

Somehow this doesn’t seem right.