Upgrading Transport Protocols using Untrusted Mobile Code (3)

From: Kate Everitt (everitt@cs.washington.edu)
Date: Mon Nov 15 2004 - 08:58:07 PST

  • Next message: Andrew R Putnam: "Review of STP"

    (I've been sending this in since 7:55. I'm sorry if there are duplicates)

    Upgrading Transport Protocols using Untrusted Mobile
    Code
    Review: Katherine Everitt

    This paper presents STP, a system for communicating
    mobile code between endpoints.
    The strongest point about this paper was that it
    addresses an important problem in networks research.
    They describe a broad solution which has several
    valuable properties: it will support the spread of new
    protocols without requiring the routers to be
    upgraded, (which is much faster), it will allow both
    endpoints to be upgraded (because this is sometimes
    necessary) and it will allow various endpoints to use
    the protocol that is best for their traffic, not
    necessarily a general solution. The most important
    issue addressed by the paper is maintaining security
    for mobile code by designing a safe API which will not
    allow malicious code to harm the kernel or take over
    network bandwidth. They use a type safe language,
    Cyclone, to achieve memory safety.

    However, there are several disadvantages that need to
    be taken into account. For example, this algorithm
    will complicate router code and provide more overhead
    for nodes. Also, this solution is good because it
    works under current conditions, but some current or
    future network protocols might not be spreadable by
    this method, and might cause problems for this
    solution. One example is a wireless protocols like
    WTCP that don’t treat loss as congestion. Other
    protocols may provide more access than STP requires,
    and this has the effect of limiting the protocols we
    can use to those who use STP’s API, which could skew
    the protocols being tried out. This solution also
    require developers to do quite a bit of work to take
    advantage of the mobile code approach.

    This paper is very relevant to today, as it addresses
    a real problem and provides a clever solution that
    allows us to test out new solutions.


  • Next message: Andrew R Putnam: "Review of STP"

    This archive was generated by hypermail 2.1.6 : Mon Nov 15 2004 - 08:58:13 PST