The Future of the Internet And How to Stop It

 

by Jonathan Zittrain

Keeping the Net Stupid

A review by Hal Abelson

The spectacular achievement of the Internet is a success that has many parents. But when it comes to engineering design, a top honor must go to the decision to make the Net "stupid": Let the network perform its limited function of transmitting bits, and leave "smarter" functions, such as encryption, content filtering and quality of service, to be supplied by the computers attached to the network rather than by the network core itself. In other words, let the network do its basic job while staying out of the way of everything else. In 1984, three designers of the Internet communications protocols -- Jerome H. Saltzer, David P. Reed and David D. Clark -- published a paper in ACM Transactions on Computer Systems in which they dub this approach the end-to-end principle. It spawned a communications system of enormous flexibility, one that was able over the course of a quarter-century of mind-boggling innovation and growth to adapt to accommodate numerous new devices and applications.

The end-to-end principle demonstrates the Internet designers' good sense, and their humility, in appreciating that they could not possibly predict in the early 1970s all the things people might want to use the Net for. They chose therefore to restrict those potential uses as little as possible. End-to-end as an engine of innovation has become a watchword in communications policy as well as technology (see, for example, Lawrence Lessig's The Future of Ideas: The Fate of the Commons in a Connected World [2001]). And end-to-end arguments have permeated much of the wrangling over "network neutrality" over the past year as advocates of network neutrality have appeared before Congress and the Federal Communications Commission.

Yet end-to-end has a dark side. After all, if you can use the Net for anything, then you can use it for anything -- including spam, denial-of-service attacks and computer break-ins carried out by spoofing IP addresses or poisoning domain-name server caches, all of which are enabled by the simplicity of the Internet's core architecture.

http://www.powells.com/review/2008_11_30.html