Kernel really necessary?
Thu, 8 May 1997 12:13:45 -0500
| From: Chris Hanson <email@example.com>
| So, is a "kernel" of sorts really necessary for a Lisp-based OS?
| What I'm thinking is really needed is:
| * A Lisp compiler that can produce a self-contained output image
| * A bootstrap that can load & run such an image from a raw partition
| * A set of primitives that can be used to access the MMU, I/O ports, etc.
| * A device driver architecture
| * Basic console and IDE device drivers
| Sure, there'd be no filesystem to begin with, but the initial OS wouldn't
| be self-hosting. However, it would be something that could be
| bootstrapped, refined, and experimented with quickly.
What you just described is almost exactly the Flux OStoolkit, which I am
studying. By the way, when we use the term "kernel" in reference to the
LispOS what are we really referring to? To me "kernel" just means a set
of functions and objects that are used in common by all the higher-level
layers of the LispOS and which should never be swapped out to disk, for
performance reasons if nothing else. They wouldn't be a static lump of
code of course. Is there a more precise - and concise - name for such an