Two events happened on Friday, September 25th, in Washington, D. C. that included lots of interesting discussion around the issues of network management and the evolution of the Internet, and broadband competition. I want to cover both but let me start with the event where the “end-to-end” principle was discussed. I’ll do a second post later this week that summarizes the second event.
The Information Technology and Innovation Foundation held a session to release a new paper by Richard Bennett, a long time engineer involved with the development of network protocols for the Internet, entitled “Designed for Change: “End-to-End Arguments, Internet Innovation, and the Net Neutrality Debate.” Appearing along with Richard was John Day, a professor of computer science at Boston University Metropolitan College and a long-time participant in Internet engineering activities, Christopher Yoo, a professor at the University of Pennsylvania, Dave Farber, a professor of computer science at Carnegie Mellon University often called the “Grandfather of the Internet,” and Bill Lehr, a research associate in the Computer Science and Artificial Intelligence Laboratory at MIT.
Overall, the commenters at the ITIF forum offered two basic points in my view: the Internet is facing many challenges, needs to evolve, and network management is an important part of the process of change; and government rules or constraints could well interfere with important changes and technology approaches that will need to be adopted to ensure the Internet can evolve as needed to meet the challenges. Assuming the “ends” of the Internet are where all changes or improvements are best implemented is not the right way to look at its future.
Here are some of the key points made by the presenters. I also include a link to the recorded web cast that you can watch.
Richard Bennett said that some of the claims made in the net neutrality debate are not consistent with how the networks actually work. He noted that advocates have used the “end-to-end” principle that was first articulated in the early 1980’s by David Clark and David Reed among others and have taken a principle based on change and unintentionally turned it into a principle hostile to change. He said end-to-end is a good principle, but it does not actually imply that there is no intelligence within the network. Instead, it was originally based on the notion that the Internet was an “experimental” technology and would have to continually change. Connecting networks involved lots of routers and servers and many network links and experimenting with changes or ideas in the “core” would be hard and problematic. Instead, experimenting on the “edges” was easier but it was also felt that if it made more sense to implement changes in the networks or the core, that was not forbidden. More “smarts” in the core, in other words, was not considered to be a bad thing.
Richard went on to say that there are design problems in the internet that have been deferred because Moore’s Law allowed us to neglect them. We’ve been able to avoid some problems because processing speeds have continued to go up. With today’s challenges of ever more complex content and security issues among other things, that is no longer going to be possible.
Following Richard’s comments, the other experts took part in a panel to discuss his paper. Here are some highlights from their comments.
John Day said that Richard Bennett was right about the internet being designed ad hoc. He said it has worked, but it has allowed flaws to persist and those will need to be addressed. Dr. Day said that it’s not that end-to-end is wrong but that it misses the point. There is no “end”. There’s no such thing as a “dumb network”. The “dumb network” is a myth. End-to-end allows only a flat network, raising barriers to solutions that scale.
Christopher Yoo said that network design questions are fundamentally about engineering. He said that Richard Bennett’s paper shows the flaws with simplistic visions of end-to-end and the importance of preserving room for innovation. According to Yoo, engineers who developed the internet believed end-to-end was not sufficient and we would need intelligence in the network. Christopher also had some slides that he released at the ITIF event.
Bill Lehr said Richard Bennett was a bit unfair to the 1981 paper on the end-to-end concept. He said these network issues are multi-disciplinary, and require input from a lot of fields that don’t often know how to talk to each other. Lehr said there is a great deal of debate about where intelligence should primarily reside – at the edges or in the network – but it’s a false choice. It’s contextual. The debate over who has the right to change or control functionality is important. It is not a dogmatic decision or issue.
Dave Farber spoke via video webcam. Dave said that the development of the Internet was more about trying to make it work, not trying to make it secure or ideal. The internet was built with the assumption of a “nice community” in the early days. As it has grown, it has encountered more malicious participants. There are serious challenges we are facing today that need to be addressed and we don’t have all the answers. We need to deal with security and how to make video work. David said those listening should see his 2007 op-ed in the Washington Post about net neutrality. He said we have to create an environment where innovation is possible. He went on that the marketplace determines what is acceptable or not, and so far that has gotten us a long way. Dave concluded that he is not a believer that the market is always right, but after being around Washington, DC for many years, I’m semi-convinced that I’d rather try the marketplace than the regulatory environment.
Interestingly, today the Washington Post has an editorial entitled “The FCC's Heavy Hand; Federal Regulators Should Not Be Telling Internet Service Providers How To Run Their Businesses”. It makes some of the points the commenters’ made but adds an additional thought – that transparency regarding network management practices is a useful idea. The FCC Chairman referenced such a proposal in his speech of last week and I have commented on this previously in a post. I do think it is a worthwhile idea, although it can and should be worked out by the industry in my view.