Review of Dennis and Van Horn's Programming Semantics paper...

From: Gail Rahn (gail_at_screaminggeek.com)
Date: Mon Jan 12 2004 - 18:14:19 PST

  • Next message: Honghai Liu: "Review of Programming Semactics for Multiprogramed Computations"

    The "Programming Sematics for Multiprogrammed Computations" paper by Dennis
    and Van Horn describes meta-instructions used in programs that are run in
    multiprogramming environments. The paper describes a low-level grammar,
    between assembly and algebraic languages, that handles privilege-granting to
    "computational objects" (processes, etc.). The meta-instructions handle
    granting and revoking read/write/execute privileges for data blocks, process
    creation and termination (fork, quit and join), privilege inheritance,
    mutual exclusion, debugging. The authors describe a toolbox of objects that
    are useful for programmers in parallel/multiprogramming situations. The
    authors also create interesting supporting notes in the paper, including an
    early aside about the parallel program design features.

    The authors propose a C-List as the unit of privilege for a particular
    segment. A c-list is list of segments adn the privileges assigned to it,
    relative to the process owning the C-list. The C-list is created by the
    supervisor program at runtime.

    The authors indicate that the described system is necessarily incomplete,
    because the expectations of operating system support of running processes is
    evolving and more features will likely be considered standard in the future.

    I especially enjoyed the paper's discussion of protected entry points. It
    sounded ot me like an early discussion of shared libraries or DLLs, where
    only certain entry points are allowed into a separate blob of executable
    code.

    The discussion of directories and naming was confusing to me. I found it an
    incomplete evaluation of the importance of string names for files. Also, the
    idea that "people are used to thinking in terms of invariant names" is
    clearly wrong and outdated. If I name a file "last month's expenses" then in
    the next month, I am expecting to rename it.
    Also, I didn't find the idea that a file's privileges are somehow based on
    its file path, whether the file exists in a principal's root directory, to
    be particularly relevant in the modern age.

    -- Gail.
    grahn_at_cs.washington.edu

    -------------
    Gail Rahn
    gail_at_screaminggeek.com
    206.719.5563

    Screaming Geek Software
    www.screaminggeek.com


  • Next message: Honghai Liu: "Review of Programming Semactics for Multiprogramed Computations"

    This archive was generated by hypermail 2.1.6 : Mon Jan 12 2004 - 18:14:27 PST