Using CL with Linux or Hurd

[Jay Lepreau]

Someone pointed me to this new list and interest and effort-- sounds
cool.  I am the head of the Flux project at the University of Utah; as
one might imaginge, there are a lot of ex-Lisp people around here, and
we've certainly thought of an oskit-based "PC Lisp machine," though we
have no intention of doing so ourselves.

I am not sure of the best OS route for you to take, but I do want to
clear up an important misunderstanding about the Flux OSKit.

	>1) Flux toolkit - most people believe this is too
	>   ambitious, and I agree.  It would be different it
	>   this was a funded project with full-time programmers.

I don't think it's so hard.  Let me give you some anecdotes:

1. Without anyone's prior knowledge, one afternoon in February one of
our students decided to try to port the free Kaffe JVM to the OSKit.
By 6am the next morning he had simple Java applets running!  Soon the
JIT compiler worked and some weeks later, complicated multithreaded
programs like the JigSaw Web server.

This guy knew Kaffe well and knows the PC fairly well.

2. A few days later in February, another Utah student who does not
know the PC well nor the oskit, but loves Smalltalk, sent this message:

    GNU Smalltalk now boots on oskit, bringing the number of languages
    running directly on oskit (that I know of) to four.  And boy is it
    cool!

    Total elapsed time (not counting sleep and food): 7 hours

    Of those 7, 2 were wasted by me not realizing that the thing wasn't
    crashed, write(1,...) just wasn't implemented, and another 30 mins was
    figuring out how threads work in GNU's wacky dialect...  ;-)

    Given the library of stuff I've collected for support, doing it again
    would take all of 20 minutes.  ;-)

3. A visiting student from UC Davis has the SR parallel language
running directly on the PC hardware, using the oskit.

4. Late last year, the ML effort at MIT took longer.  This is partly
because it was tackled by a student who was new to the PC, partly
because they never asked us for any help, and partly because the OSKit
was less mature then.  However, mostly it was because he was new to
the very complex ML runtime.

5.  People here have threatened to port emacs to the oskit, making an
emacs machine, but we have been too busy.  I bet this would be easy
and a fun demo!

These were all essentially one or two person efforts, and some took a
matter of hours, some days or weeks.

---

Awhile ago I withdrew the previous release and a new release is due
out in mid-May.  Internally, the OSKit is now lots more functional.
Currently we have most of a "Java computer" running, because since
February we have added new libraries providing major components like
socket based networking from FreeBSD, the NetBSD ffs filesystem, and
the x-kernel network protocol framework.  Our device support is now
more flexible and sophisticated.  A simple, nearly pthreads-compatible
threads library is almost done.  The major missing thing today is
display support, but we are making progress there, working on
providing X, which should be available in May.  Eventually we will
have a lighter-weight display system, but are not sure what we will
base it on.

I am also willing to commit to help you guys by having our experts
here answer questions, and, other things being equal, to slant Utah
development efforts in directions that you need.

We have just added some pretty pictures and outline of the components
in a Java-oriented page hanging off the oskit page:
http://www.cs.utah.edu/projects/flux/oskit/.

Jay Lepreau, lepreau@cs.utah.edu, +1-801-581-4285.