Werbach: Forget neutrality, regulate interconnection

by on March 1, 2007

Map of internet backbone ownershipWharton professor Kevin Werbach has posted an interesting new paper on net neutrality that’s not really about net neutrality. His thesis is that while he agrees with the proponents of regulation that broadband network operators will disadvantage content and application providers and thus stifle innovation, he doesn’t think anti-discrimination rules are the way to go. In fact, he does a great job of explaining why they’re not a good idea and how discrimination of all kinds–from content delivery networks (CDNs) like Akamai, to propriety video services like ESPN 360–serve the interests of both consumers and network operators. He also highlights how difficult it would be under neutrality rules to distinguish anti-competitive discrimination from benign discrimination like spam blocking or legitimate traffic management.

Instead he argues that the real issue missed by the neutrality debate is interconnection. “The defining characteristic of the Internet is not the absence of discrimination, but a relentless commitment to interconnection,” he writes. Networks withholding interconnection is the real threat to innovation. In particular, he is concerned about access tiering, “[broadband networks] charging content and application providers additional fees for preferential access to their broadband access customers.”

Today’s Internet already countenances discrimination, in the sense that only larger backbones are entitled to peering, and only those who pay for private exchange points and CDNs receive the benefit of enhanced delivery they offer. It is also a two-sided market, in that network operators may receive revenue both from end-user customers and from content and application providers on the “server side” of the network. Neither is a challenge to the “neutral” character of the Internet, because, from the end-user perspective, there is still one universal network. Providers of content and applications that desire enhanced delivery have several options: they can buy a bigger pipe or a stricter service level agreement (SLA) from their backbone operator; they can go to a private exchange point or CDN that overlays intelligence on the Internet infrastructure; or they can self-provision distributed capacity, as companies such as Google and Microsoft do today.

If access tiering is widely practiced, such providers will have more limited options. If the backbone operator connected to a last-mile network conditions enhanced delivery on the purchase of quality of service (QOS) capabilities that it hard-wires into its network, that operator becomes the sole arbiter of how the content or application provider can reach some of its customers. Connectivity across the Internet becomes less of an opportunity to take advantage of pervasive interconnection than a set of isolated, private negotiations with broadband carriers.

Werbach lays out several other threats to universal interconnection and argues that we need interconnection mandates to remedy the situation. So far, we haven’t seen the interconnection of the internet really threatened. In a 2005 paper (PDF) cited by Werbach, Randy May and Richard Levine explain how competitive markets in the internet backbone business have led to interconnection without regulation.

The key value of an Internet connection is that it connects a user to any other Internet user. Because no backbone operator is dominant in terms of the number of systems and end users connected to it, it is in the economic interest of each backbone operator to interconnect with other backbone operators.

Werbach is unsure this will continue:

One reason to be skeptical the voluntary peering regime will endure is the consolidation of the backbone market. Following a series of mergers, a relatively small number of operators dominate backbone capacity, and three of them–AT&T, Verizon, and Qwest–are also the major incumbent telephone companies in the US. There is a perception that backbone providers originating in the data networking world are more likely to observe the Internet cultural norms that promote voluntary peering. In contrast, the dominant paradigm of the telephone world is regimented, billed settlements for all traffic. The current balance of peering and transit arrangements for the Internet backbone is not written in stone; it is a particular stage of evolution in an industry marked by constant change. If transit-type arrangements replace peering for a significant percentage of major backbone interconnects, however, the basic economics of Internet transport could shift in unpredictable ways.

However, as May and Levine explain, whenever competition in the backbone market has been threatened, antitrust regulators have acted, requiring, for example, the divestiture of backbone businesses before the MCI-Worldcom merger was approved, and the blocking of Worldcom’s acquisition of Sprint. To the extent that concentrated market power is used anti-competitively, antitrust enforcement can be used correct the problem. I’m not sure that “a perception” that the telephone-centric entrants to the backbone market have a regimented bill-everything mindset justifies prophylactic regulation that could have unintended consequences.

May and Levine also acknowledge that vertically integrated closed networks could emerge along-side, or mixed with, the wider open internet. If some networks do in fact decouple completely from the rest of the internet, we’ll just have to see how consumers react to that. Consumers might show that they value other things beside universal interconnection. Some, like Wayne Crews, have been arguing for years for just such splinternets. As long as there is competition in last mile delivery–again, an antitrust question–consumers would be unaffected. As I’ve asked before, what consumer would subscribe to an ISP that didn’t let her connect to Google or receive e-mail from her mom who is on AOL? Those who value interconnection–at both ends–will choose networks that supply it.

Comments on this entry are closed.

Previous post:

Next post: