Reading Review 11-17-2004

From: Craig M Prince (cmprince@cs.washington.edu)
Date: Tue Nov 16 2004 - 12:38:23 PST

  • Next message: Ethan Phelps-Goodman: "DNS"

    Reading Review 11-17-2004
    -------------------------
    Craig Prince

    The paper titled "Development of the Domain Name System" gives a
    historical look at the creation of a naming system for the original DARPA
    internet. The problem was the original method of passing around a single
    file with all the names was not scalable. As a result, a new method of
    resolving names into address was necessary. The main features that the
    developers thought were necessary were to not lose any of the
    functionality of the previous system, have the system be distributed (for
    reliability and performance), and not to place artificial constraints on
    the system.

    Basically, the system that they came up with has a hierarchy of servers to
    resolve names. The naming convention they chose to follow was the
    cannonical '.' delimited name strings (with some special cases for e-mail
    addresses). The lookup then progesses using the suffixes of the name.

    One thing I especially liked about the paper was that it focused a lot on
    the performance details of the system. Namely caching and how hierarchy
    helps scale the system without losing performance. I also liked how the
    paper explicitly looks at the "shortcomings" of the system.

    A concern I have with the system is that in general, the internet was
    designed to be reliable even under faiures; however, the domain name
    system, seems to not have this property. The root servers themselves are
    points of failure. There is redundancy built in; however, if these servers
    go down, then we either have to rely on cached copies or we're stuck. A
    more distributed system could perhaps solve this problem or at least be
    more efficient.

    Overall, I thought this paper provided interesting insight into how the
    naming system was forced to evolve as the size of the network increased.
    However, in this case the DARPA internet was still fairly young and so it
    was much easier to make these sweeping changes. It is interesting to
    consider how one might overhaul the naming system in today's internet?


  • Next message: Ethan Phelps-Goodman: "DNS"

    This archive was generated by hypermail 2.1.6 : Tue Nov 16 2004 - 12:38:23 PST