A Protocol for Packet Network Intercommunication

From: Jonas Lindberg (jonaslin@kth.se)
Date: Wed Oct 06 2004 - 18:06:46 PDT

  • Next message: Kevin Wampler: "Review of Cerf and Kahn paper"

    Jonas Lindberg's review of "A Protocol for Packet Network
    Intercommunication" by Vinton C. Cerf and Robert E. Kahn:

     

    As the title indicates, this paper presents a transmission control protocol
    that enables communication between different packet switching networks. In
    this paper, Cerf and Kahn first explains how packet switched networks can
    differ and how these differences can be overcome using their suggested
    TCP-protocol and gateways. The goal of TCP is to provide reliable and
    transparent access to shared resources on the interconnected networks. In
    this paper the authors discusses a number of techniques and concepts used
    for achieving this goal, for example: addressing, buffering, sequencing,
    error control, flow control and reassembling.

     

    The paper is generally well written; the authors explain the problems in a
    way that is easy to understand, presents alternative solutions and carefully
    motivates their choices. Two of the parts that I find particularly well
    explained are Cerf and Kahn's opinion that:

    1. No maximum packet size should be specified (p. 639).
    2. Gateways should fragment packets when necessary, but they should
    never reassembly fragmented packets (p. 639).

     

    When it comes to the papers shortcomings, one thing that is obvious thirty
    years after the paper was published is that the authors could have been a
    little bit more careful about making statements like: 256 networks being
    sufficient for the foreseeable future and 65 536 TCP's seeming more than
    sufficient for any given network. I think one way to improve this paper
    would be to insert a few pictures that illustrate the third paragraph of
    "The Gateway Notion".

     

    The paper is interesting because it gives the reader an idea of the
    reasoning and thoughts behind the design of the TCP. Considering how
    important and widely used the Internet is and that TCP is one of its
    fundamental building blocks, I think this paper is very relevant.

     


  • Next message: Kevin Wampler: "Review of Cerf and Kahn paper"

    This archive was generated by hypermail 2.1.6 : Wed Oct 06 2004 - 02:07:43 PDT