Review of D. Clark, "The Design Philosophy of the DARPA Internet Protocols"

From: Kate Everitt (everitt@cs.washington.edu)
Date: Mon Oct 04 2004 - 09:44:30 PDT

  • Next message: Pravin Bhat: "Review2-PravinBhat"

    The Design Philosophy of the DARPA Internet

    David D. Clark. MIT

    Synopsis: Katherine Everitt

     <>

    This paper discusses the goals involved in the design of the internet,
    some design decisions based on these goals, and the success and failure
    of the design decisions at supporting the goals. This is a very
    important problem, as evidenced by internet use today. The protocol
    authors discussed in the paper succeeded at their main goals quite well,
    but found some of their decisions were not ideal in supporting some of
    their secondary goals and some practical aspects of actual use.

    The main design decision discussed in the paper was the use of a
    datagram model over a packet switched network. The other main design
    decision was to split TCP and IP, as TCP was too specific to support all
    the needed types of service.

    One of the most interesting things I felt about this paper was the
    disparity between the design goals and the actual user goals. One of the
    most obvious ones was the issue of trust of hosts. An unstated
    assumption was that the hosts were trustworthy and part of the internal
    model but in practice the Internet is a system where a host can
    misbehave and hurt the network. This problem is still seen in the usage
    of the internet today.

    This opens up the question of how one designs a system that is flexible
    enough to support shifting goals. This was partially addressed in the
    use of a datagram network and the IP protocol. From this basic building
    block, one can develop more sophisticated services which could give
    different models of reliability, delay and bandwidth.

    In general this paper helps in understanding the reasoning behind the
    TCP/IP protocol, and is very valuable when proposing and evaluating
    changes or the creation of similar protocols. It would be more effective
    if it included a modified list of main goals based on what was learned
    during the development of the protocol.


  • Next message: Pravin Bhat: "Review2-PravinBhat"

    This archive was generated by hypermail 2.1.6 : Mon Oct 04 2004 - 09:44:41 PDT