Our Manifesto (Please Read)
Wed, 7 May 1997 08:40:10 -0700
First of all, let me thank you for the work you've put in. Having played a
similar role in the past for other efforts, I know how hot the editor's seat
can be. <g>
Most of this is fine, and LispOS is my vote for the name -- why get cute?
I would like to comment about the LispVM project mention:
1. Bytecodes have not been agreed with. I favor wordcodes for efficiency
reasons, and because we have a working wordcode instruction set that is already
efficient for Lisp, Scheme, etc. I'm reviewing Kaffe as a code base to start
with, but that may be to brutal -- when Mike & Gilda (Mike's wife and compiler
wizard) get back from Brazil, I want to get Gilda's suggestions on modifying it
before I get going.
2. My opinion of the primary goal of LispVM is to produce a remoteable,
cross-platform virtual machine that is optimized for Lisp and Scheme and is
binary-compatible with LispOS. To be more specific, a VM that can be embedded
in Web-servers, browsers, or just accessed as a DLL from any application. I see
this as providing LispOS apps with some level of portability and
interoperability -- goals I believe are very important.
3. Starting with UNIX? Well, I'm going to get Linux and CMUCL, but that has to
wait for a machine to arrive, etc. I'm probably gonna start under NT and use
the Linux box to test portability (and learn Linux as an atonement for past
I think that, once LispVM is done, LispOS will want to embed it. By then folks
will have realized that they can't easily put Lisp on top of Scheme, or vice
versa, and that to do both justice, you need a VM that can accomodate both with
equal efficiency. Then, to satisfy the speed mongers, a VM->native translator
(or JIT) should do well.
My 2 cents. YMMV
The views expressed are mine alone,
unless you agree with me.