Paper review: The Design Philosophy of the DARPA Internet Protocols

Reviewer: Hanlin D. Qian

In this paper David Clark addresses the problem of how to connect existing computer networks into a large network now known as the Internet. This problem goes to the heart of all the research done on the Internet. This paper presents one solution to the problem, based on a certain set of assumptions as to how the Internet should work. This solution is actually how the Internet is currently implemented. Although the solution is generally considered a success, it is based on a philosophy that made sense back in the 1970s. The Internet was originally designed for military defense purposes with a set of assumptions that have less practical use for today's Internet, which has moved toward E-commerce and the globalization of communication in many media.

Clark's paper focused on three main ideas behind the philosophy of the design of the Internet. First, it states that the designed emphasized reliability, flexibility, and the support of multiple networks and services over cost efficiency and accountability. As a result, we now have an Internet the is not a centralized entity but a decentralized and distributed collection of networks. I like to think of as a combination of several nation states without a centralized government. Second, the design of the Internet is layered. What this means is that TCP and UDP are on a separate layer as IP. This design allows the bottom layer, IP, to be kept as simple and flexible as possible. Third, this paper emphasizes on simplicity over complexity. As a result, instead of having mechanisms in IP make sure packets are delivered correctly, the burden of verification is placed on the end hosts, or what the author calls as "fate-sharing." All of these design criteria led to a highly decentralized architecture that is reliable but not as accountable.

This paper is a breakthrough in the area of Internet design. It offers a solution that is highly scalable, which is why the number of hosts on the Internet has grown exponentially in the past decade. Even in today's world, even though some of the original assumptions behind the design no longer applies, the solution presented by the paper is still considered highly successful. The Internet does have problems, but from a design point of view, simplicity and scalability are two important aspects of the Internet that made it successful, and the originally design made it so the both were possible.