"Programming Semantics for Multiprogrammed Computations." Review

From: Tarik Nesh-Nash (tarikn_at_microsoft.com)
Date: Mon Jan 12 2004 - 09:23:49 PST

  • Next message: Greg Green: "Programming Semantics For Multiprogrammed Computations"

    Many different sequential Computations were implemented in the late 60s,
    and many high level languages were already available . This paper
    raises the question of multiprogramming systems and explains how the
    existing technology does not provide any good support. The author then
    proposes an intermediate language that addresses some of the aspects of
    multiprogramming, with the expectation that high level languages will be
    extended accordingly and this intermediate language will grow as needed.
    I m a bit not clear about the distinction between multiprogramming and
    time-sharing systems. I would describe a multiprogramming system is
    based on machine resources sharing by multiple concurrent applications,
    while time-sharing is based on the idea that multiple users are using
    the system. The definition seems different but tightly related.
    The document describes five properties that do not exist in conventional
    systems and that the research tries to address. A careful definition of
    the terminology was detailed, and this makes the reading much easier and
    avoided any discrepancy.
    A definition of different primitives were introduced to solve the
    problems of Parallel processing, naming and protection. This looks like
    a pioneer paper on this subject. Some solutions were still basic and
    had some architectural holes.
    For example, the parallel processing implementation had the definition
    of lock/unlock and avoided starvation using timeout, however, since the
    operation is not atomic there is a potential of data corruption.
    Another unclear point is about protection, it seems that the system
    trusts the user, the user can set any rights to its process, and hence
    there is no protection against malicious users (this may be irrelevant
    to the use of this particualr system).
    Overall, this is a very entretaining paper, sometimes criptic, sit
    should be an original and innovative publication on the time it was
    published.


  • Next message: Greg Green: "Programming Semantics For Multiprogrammed Computations"

    This archive was generated by hypermail 2.1.6 : Mon Jan 12 2004 - 09:24:20 PST