6.7 The importance of end to end

The end-to-end (e2e) design was first described by David P. Reed, Jerome H. Saltzer and David D. Clark in 1981. Over 20 years on they still defend the importance of end to end as a design that leads to a more flexible and scalable architecture. The principle is to keep the intelligence at the ends of the network and keep the network itself simple.

We have already seen some of the mechanics of how this works. We have also seen that architecture matters. Robert Moses' low bridges regulated social interactions between different ethnic groups whether he intended them to or not.

Internet architecture matters. An e2e design means that:

  • Innovations only need to connect to the network to run. Permission from a network owner or other central authority is not required – any more than permission would be required from the electricity company to plug in a new kind of electrical device.

  • The network owner cannot discriminate against specific data packets, slowing some down and speeding others up, because the network is simple. It doesn't provide the facility to discriminate. It shunts packets from point to point roughly in the direction of their ultimate destination, until all the packets (or their replacements in the case of those that don't make it) arrive and get reassembled into the original file.

  • As David Reed has said, e2e does not ‘restrict our using some new underlying transport technology that turned out to be good in the future.’ Since e2e is not optimised for any existing use the network is open to innovation not originally conceived of, such as the World Wide Web.

End to end is a design choice. The original network architects chose it specifically because they did not want to restrict possible future uses. As Lessig says, the network would not control how it would grow, but innovators with new applications would. Tim Berners-Lee has also said of the World Wide Web that to be successful it needed to grow in an unlimited way. Allowing some form of central control would have led to a bottleneck as the Web got bigger.

The telephone network is optimised for phone use. It is an intelligent network to that extent. It also means that it can't easily be used for new applications not originally thought of because changes in one area have all sorts of knock-on effects on the system. The internet grew on the telephone wires. The physical layer was controlled. The telephone companies were prevented by law, though, from controlling internet applications run over their networks. Hence the end-to-end disabling of control was extended into the physical layer, in this case by law.

Lessig argues that e2e at the code layer makes the internet an innovation commons. It allows a broad range of development and facilitates the disruptive innovation we talked about in Section 2. Established interests, commercial or political, are not good at facilitating disruptive innovation, and concentrated control does not often produce disruptive technologies or ideas. End to end provides a route around established interests for creators and innovators.

"The person who says it can't be done should not interupt the person doing it."

(Chinese proverb)

6.6.2 How a stack works

6.8 Activity 5