thinking about virtual machines
Sun, 27 Apr 1997 23:34:34 +0200 (MET DST)
>: Michael Korns
you seem very enthusiastic about a VM design.
Now, I'd like that you answer my previous post on the list,
and explain what are the goals and constraints of your VM,
and what general kind of VM design you envision,
and what in this design is justified by the goal,
what comes from what other reasons.
For instance, I'd like to hear why the LISP core language
isn't enough of a VM to you, etc.
> First, we should develop an "LVM" standard "straw dog" and quickly develop
> code around it.
What is there to develop? You're focusing on implementational issues,
but we still do not know what to code with this implementation technique,
so how can we know whether it's fit for the goal!?
Again, you seam to imply that the VM would specifically emulate
a von Neuman machine, much like current hardware CPUs.
why ain't (a suitable encoding of) the core LISP language satisfying?
Until we have it, we can already program in LISP (which, btw?)
(any EuLisper around?).
> Second, we must start with the Java VM standard and extend (not alter it).
Foo! If by chance the JVM has room for extension, then be sure that *they*
(or Microsoft) will extend it, so don't you expect stay compatible if you
do more than strictly following (unless you can outhype the Java people);
if you want JVM compatibility, your only chance is
to pressure the mainstream JVM to include your extensions,
of which you would give a prototype.
This might be possible -- after all, Guy Steele is co-heading Java design;
but this isn't the main purpose of the lispOS project, afair.
> It can be done quickly (say six months to critical code mass).
> It is up to the LispOS community to decide.
If you want that, you'd better go ahead and do it;
you seem to know better what you want that we do.
I might be wrong, but I don't think you can expect more in your project than
helpful suggestions from this project, especially in a six-month delay.
Your project to do a better JVM as the basis of a standardization
of your commercial Agentbase system seems worthwhile
(except that I personally only trust free software),
but it is clearly distinct from the lispOS project.
PS: it's Wrong(tm) to entirely quote the message you answer to
at the end of yours.
== Fare' -- firstname.lastname@example.org -- Franc,ois-Rene' Rideau -- DDa(.ng-Vu~ Ba^n ==
Join the TUNES project for a computing system based on computing freedom !
TUNES is a Useful, Not Expedient System