Packet Network...

From: Ethan Phelps-Goodman (ethanpg@cs.washington.edu)
Date: Tue Oct 05 2004 - 22:04:37 PDT

  • Next message: Michelle Liu: "Review of "A Protocol for Packet Network Intercommunication""

    A Protocol for Packet Network Intercommunication
    Cerf and Kahn

    Cerf and Kahn present a high level specification of the TCP/IP
    protocol. The protocols are specifically aimed at linking existing
    heterogeneous networks. They enumerate potential hurdles for connecting
    disparate networks: varying addressing schemes, packet size and delay;
    reconstructing lost data; and monitoring network state. As mentioned in
    the previous paper, the last goal is largely unmet.

    Networks are linked by a gateway that sits on two different networks at
    once. This seems to be a new contribution of this paper. The IP layer
    requires very little of the underlying networks. The underlying
    networks may use any packet size, routing scheme, and header
    information. IP is responsible for specifying a global network
    identifier, but the individual network decides how to interpret the
    host field, and the individual host decides how to interpret the port
    number. A lot of attention is paid to sequencing and reconstruction,
    retransmission, and managing connections (a term which they somewhat
    dislike.) A key strength in all of this is that the fundamental design
    is made explicit, but many low level details are left up to the
    implementation. This was probably appropriate since the protocol was
    still in its early stages.

    Although there are justifications for some of the design decisions,
    there is no attempt made to asses the workability of the protocols. The
    paper very much needs a followup study to assessing the performance of
    a prototype.

    Ethan


  • Next message: Michelle Liu: "Review of "A Protocol for Packet Network Intercommunication""

    This archive was generated by hypermail 2.1.6 : Tue Oct 05 2004 - 22:04:44 PDT