From: Michelle Liu (liujing@cs.washington.edu)
Date: Sun Oct 03 2004 - 21:36:38 PDT
Review for "The Design Philosophy of the DARPA Internet Protocols"
Jing Liu
This paper addresses the motivations for the Internet protocols in early time and the reasoning behind the design decisions.
The Internet has evolved for decades and is proved to be very successful. Internet protocols, such as TCP/IP are well-known today and there have been numerous papers and specifications that describe how the protocols work. However, the question why the protocols are designed in this way might be easily ignored. In this paper, the author answers this question by listing the desired goals of original DARPA Internet in order of importance. The priority of the goals guides the designing or Internet protocols.
Reasons for some fundamental designing features of the protocols are stated in the paper. For example, the reason that packet switching was accepted instead of circuit switching in the Internet architecture is to fulfill the top goal of DARPA Internet¾ multiplexed utilization of existing networks. Furthermore, the second level goals led to the design of Transmission Control Protocol (TCP), the separation of Internet Protocol (IP) from TCP and the creation of User Datagram Protocol (UDP). Thus, from this paper, we can know very clearly how those different protocols were created and the reason why they were created in this way not the others. The author does not only stop in describing the existing Internet protocols, but also shows the imperfect parts of the Internet architecture and the goals that current protocols haven't reached yet. For instance, the Internet architecture is not cost effective and resource management and accountability is still unsolved. In addition, the extremely challenging relationship between architecture and performance is stated in the paper.
The author also gives some suggestion on protocol design to achieve more goals. "Flow" is introduced to characterize the data block which makes resource management and accountability possible. The concept "soft state" is declared which is to achieve the goals of survivability and flexibility.
One of the weaknesses of the paper is that the author only addresses those goals or desired features to be the motivations of the design of TCP/IP. It doesn't mention the impact of network layering and the relations between network layering and protocols. It would also be nice if the author could explain the designing features of TCP/IP in more detail. However, it is a bit hard in this paper, since it covers quite broad issues of network design.
The paper provides guidance for network protocol designers and shows the direction of designing. In the future, new building blocks could be explored instead of using datagrams as the entity transported across the underlying Internet and thus, protocols might be able to deal with the issue of resource management and accountability or the other unachieved goals. The other possible work is to consider designing formal tools for describing performance or exploring the quality of service property of the Internet.
This archive was generated by hypermail 2.1.6 : Sun Oct 03 2004 - 21:36:49 PDT