Dijkstra Review

From: Brian Milnes (brianmilnes_at_qwest.net)
Date: Wed Jan 14 2004 - 09:40:53 PST

  • Next message: Greg Green: "Review of THE"

    > The Structure of the "THE"-Multiprogramming System - Edsgar W. Dijkstra
    >
    >
    >
    > The author describes the principles and construction of a
    > multi-programming system. The principles are: stick to the interesting
    bits,
    > pick a solid machine and attempt to learn from your experience. The
    software
    > engineering experiences mentioned is the loss of time using part time
    labor
    > and a recommendation to use very skilled labor as OS design is difficult.
    > The system was designed for what seems to be batch program execution. The
    > author laments two design errors: handling error cases in the machine's
    > operation and peripherals and design for debug ability. They achieved a
    > pretty remarkably low error rate of 1 instruction/500 instructions written
    > detected by inspection and extrapolate this to a very ridiculously high
    > correctness statement.
    >
    >
    >
    > This system appears to have been the invention of memory pages
    > backed up by slower (drum) storage. They have also arranged all OS
    > functionality such as device handling with sequential processes arranged
    in
    > a strict order and communicating using shared memory and the
    synchronization
    > primitives P and V. The hierarchy is process control, memory allocation,
    > message processing which handles sequential character IO to and from the
    > teletypes, peripheral processing and finally user programs. The system
    was
    > tested level by level and could even allow testing to proceed when the
    drum
    > broke. However, the author makes absurd correctness claims.
    >
    >
    >
    > Finally, he describes the famous P and V semaphore
    > synchronization and sketches a proof of their correctness for
    communicating
    > sequential processes.
    >
    >
    >
    > This is a very well thought out system for such an early time.
    > The author's emphasis on software engineering issues seems quite ahead of
    > its time and the P and V semaphores are a basic contribution to our
    science.
    >


  • Next message: Greg Green: "Review of THE"

    This archive was generated by hypermail 2.1.6 : Wed Jan 14 2004 - 09:41:03 PST