From: Steve Arnold (steve.arnold4_at_verizon.net)
Date: Sun Jan 11 2004 - 23:43:52 PST
The authors of HYDRA dispense with the idea that a system should very
hierarchical. Instead they belive that the whole thing should be a system of
objects that are all working together. After laying out their principles,
the authors begin to describe these objects. Mainly they consist of
procedures, LNS's, and processes.
Each procedure has some protection built into it. The model they present is
extensible (more than just RWX). In fact, the bulk of the paper is then
spent explaining this protection system. An object consists not only of its
data, but of capabilities -- that is, whether or not the caller is allowed
to invoke such an object.
They then go on to drill in a little bit more, briefly mentioning how other
resources are also made of objects (and it seems they were alluding to
drivers). As an example, a multi-user program description is given.
It is interesting how they quite openly dismiss much of Djiksta's work
(althouh they do give credit to him for his appendix). It seems that there
is only one level of hierarchy. It seems this paper is more abstract than
others that we have read. Their described system has many analagies to
object-oriented programming.
This paper was easier to read than others, although it did seem repetitive
in parts. They concentrated mostly on protection mechanisms, although it
seems that there would be more to the kernel in such a system (such as
process scheduling). They also didn't mention how it had been implemented.
This archive was generated by hypermail 2.1.6 : Sun Jan 11 2004 - 23:43:17 PST