End-to-end? Or end to innovation?

In what is sure to be a substantial contribution to both the technical and policy debates over Net Neutrality, Richard Bennett of the Information Technology and Innovation Foundation has written a terrific piece of technology history and forward-looking analysis. In “Designed for Change: End-to-End Arguments, Internet Innovation, and the Net Neutrality Debate,” Bennett concludes:

Arguments for freezing the Internet into a simplistic regulatory straightjacket often have a distinctly emotional character that frequently borders on manipulation.

The Internet is a wonderful system. It represents a new standard of global cooperation and enables forms of interaction never before possible. Thanks to the Internet, societies around the world reap the benefits of access to information, opportunities for collaboration, and modes of communication that weren’t conceivable to the public a few years ago. It’s such a wonderful system that we have to strive very hard not to make it into a fetish object, imbued with magical powers and beyond the realm of dispassionate analysis, criticism, and improvement.

At the end of the day, the Internet is simply a machine. It was built the way it was largely by a series of accidents, and it could easily have evolved along completely different lines with no loss of value to the public. Instead of separating TCP from IP in the way that they did, the academics in Palo Alto who adapted the CYCLADES architecture to the ARPANET infrastructure could have taken a different tack: They could have left them combined as a single architectural unit providing different retransmission policies (a reliable TCP-like policy and an unreliable UDP-like policy) or they could have chosen a different protocol such as Watson’s Delta-t or Pouzin’s CYCLADES TS. Had the academics gone in either of these directions, we could still have a World Wide Web and all the social networks it enables, perhaps with greater resiliency.

The glue that holds the Internet together is not any particular protocol or software implementation: first and foremost, it’s the agreements between operators of Autonomous Systems to meet and share packets at Internet Exchange Centers and their willingness to work together. These agreements are slowly evolving from a blanket pact to cross boundaries with no particular regard for QoS into a richer system that may someday preserve delivery requirements on a large scale. Such agreements are entirely consistent with the structure of the IP packet, the needs of new applications, user empowerment, and “tussle.”

The Internet’s fundamental vibrancy is the sandbox created by the designers of the first datagram networks that permitted network service enhancements to be built and tested without destabilizing the network or exposing it to unnecessary hazards. We don’t fully utilize the potential of the network to rise to new challenges if we confine innovations to the sandbox instead of moving them to the parts of the network infrastructure where they can do the most good once they’re proven. The real meaning of end-to-end lies in the dynamism it bestows on the Internet by supporting innovation not just in applications but in fundamental network services. The Internet was designed for continual improvement: There is no reason not to continue down that path.

Comments are closed.