In addition, until Lessig’s testimony before the Senate Commerce Committee, many advocates argued against any differential treatment of different packets (claiming that “bits are bits”) and argued that network neutrality prevented charging more to end users who consumed more bandwidth. Lessig backed off that position, but nonetheless opposed “access tiering,” i.e., charging more to applications and content providers that demand higher-bandwidth, lower-latency services.

All of this underscores the extent which the network neutrality debate has remained something of a moving target. If the comments mean that network neutrality proponents would now agree to permit charging more to content and applications providers that use services that provide higher guaranteed QoS, I would see that as a distinctly positive development. The role that the FCC’s Policy Statement continues to play in the debate suggests that at least some network neutrality proponents continue to be concerned by network architectures that allows networks to prioritize certain applications over others. To the extent that is the case, I still believe that preventing such prioritization would prevent the network from moving forward.

I continue to believe that Akamai represents much more than just another service riding on top of the network. Queries that are served by Akamai are handled in a very different way than other queries and with a much tighter integration into the network than is typically the case with other Internet-based services. At a minimum, placing intelligence in the middle of the network to rewrite and redirect queries is a violation of the end-to-end argument. At most, particularly as Akamai increasingly distributes content through private connections rather than through the public Internet, it can be regarded as an alternative network architecture.

Those interested might want to look at an excellent article on so-called “overlay networks” like Akamai by David Clark (one of the original proponents of the end-to-end argument) along with several other colleagues in MIT Computer Science Department. David points out that the Internet began as an overlay riding on the top of a voice network. Over time, the notion of what was “the network” and what was “the application” inverted to the point where voice is just an application riding on a data network. This paper finds it conceivable that a distributed approach might ultimately displace the end-to-end architecture that is implicit in most network neutrality proposals.

I’m not sure that any experts in private industry or in the government can foresee whether that transformation will occur. I am concerned that imposing regulation would necessarily lock a particular vision of how functions are distributed into place in ways that would survive longer changes in technology made it obsolete. Even the most well-meaning government would have a hard time keeping up with the current pace of technological change. Under the best of circumstances administrative processes are subject to delay; under the worst of circumstances, they are influenced by lobbying by private interests and agency capture.

I recognized that I should have said that “some” technologists have begun to suggest that TCP/IP is becoming obsolete rather than attributing that position to “most” technologists in my original post. I had in mind MIT computer scientist David Clark, whose work I mention above, along with Carnegie Mellon computer scientists David Farber, who some have called the grandfather of the Internet.

Interestingly, my point doesn’t really rest on whether or not TCP/IP remains the best protocol. If the dire predictions that the 2001 AOL-Time Warner merger represented the end of history have taught us anything, it is to be humble about our ability to predict with certainty the optimal network architecture or how end user needs will change in the future.

My argument is thus not in favor or against any particular protocol or architecture, but rather in favor of a regulatory approach that gives networks the latitude to explore different architectures, bearing in mind that powerful forces already exist in favor of maintaining a high degree of compatibility. (Somewhat ironically, the literature on the economics of networks focuses on “tipping” that may cause networks to become “locked in.” This in turn suggests that network architectures may be too reluctant as well as too willing to change.)

We should thus be wary about imposing any regulation that would systematically privileges innovation at the edge over innovation in the core of the network or pre-commits to any particular architectural vision. Such regulations would threaten to choke of the room for experimentation upon which the process of technological progress depends.