From: Steve Arnold (steve.arnold4_at_verizon.net)
Date: Sun Jan 11 2004 - 14:49:44 PST
In their paper, Dennis and Van Horn lay much of the groundwork for what is
to come in what they call "Multiprogrammed Computer Systems," or MCS. They
give examples for why this is needed (although this is obvious now). They
give a brief overview of protection, which is somewhat similar to what we
see in future papers. The basics of processes, threads, and segments are
laid out, explaining how you must be able to stop and start various
processes.
Some overview is given for parallel programming, such as how to enable it,
but little is given as far a examples for how it should be used. In what
they call a supervisor, much of the overhead is taken care of for these
processes. They've also given thought to locking and unlocking of resources,
but I think more advanced topics come in subsequent papers.
They also give a lot of emphasis on "inferior spheres of protection,"
explaining how they are use for debugging and can started and stopped by a
superior process. Lastly an overview is given for how to deal with "retained
objects." They have thought a lot about naming and security of these
objects.
In this paper many of the terms are out-dated, but the underlying concepts
still apply at many levels. They never once mention an operating systems,
but you can see how that would develop out of this. Also I found the file
system to be quite primitive with its restrictive use of names (but at least
they allow this; naming was not allowed in C-lists or other programming
semantics).
The paper did take quite a while to read. Their examples are a bit hard to
follow, with the use of i, j, k instead of useful mnemonics. It would also
have been nice to see more examples of how they plan to use all their
functions, better tying things together.
This archive was generated by hypermail 2.1.6 : Sun Jan 11 2004 - 14:49:04 PST