[eros-arch] Mothballing EROS

[Jonathan S. Shapiro]

This note is going to upset a lot of people. I hope that you will read
it all the way to the end and *think* about it before replying.


I have decided to mothball the EROS project for at least a year. After
the work of the last two years, I have reached the conclusion that there
are some important failings in the EROS/KeyKOS architecture. There are
also many minor irritations, but this is true of *every* operating
system. I will describe the issues in subsequent notes in the interest
of generating discussion. I want to spend the next year or two trying to
find a hybrid design that can achieve an effective architectural balance
-- preserving the things that EROS got right and fixing the two or three
things that I think are critically wrong.

It may prove that this hybrid does not exist. If not, I think this is
something that we as a community and I as a researcher need to know. My
opinion is that it *may* exist -- I have been background processing
ideas over the last eight months, and some of them look quite promising.
I'll be capturing some of them in writing over the next few weeks. There
are two points at which the hybrid work can be killed early if it is
failing:  at the design stage, or about 12 months into the
implementation when we get a first working system. If it turns out that
the hybrid ideas don't work I plan to return to the EROS effort.

Before anybody asks, we will update the tarball so that people can get
it, and we will continue to host the repository for others who may want
to keep working on EROS. As soon as OpenCM 1.0 goes live (should be
shortly), I'll set up a community branch if anyone expresses interest.
The current web site is drawn automatically from the OpenCM tree. If the
project goes community I'll update the scripts to update it from the
community branch.

Before we do anything too drastic, though, let's start talking about the
issues with EROS in the next few email threads.

shap
-- 
Jonathan S. Shapiro <shap@eros-os.org>