Fri, 27 Mar 1998 10:11:07 -0500 (EST)
> Yeah. That would be great. If LispOS is based on a VM, then
> the task gets easier (I hope). If tcl/tk, python and java are
> cross-platform (even with GUI's), then LispOS should also be. The only
> drawback is speed.
> The question is: is it still worthy a fairly complete LispOS
> system ontop a virtual machine? Is it fast enought?
> How complicated is to implement a JIT virtual machine for
Lisp is already a virtual machine in most respects, its types are
definable independant of the underlying architecture, it defines
a machine based on loose standards in at least most useful cases.
If you merely specify a lisp system sufficiently well that it is
not ambigious in definition, and such that the resulting system
does not reflect the underlying system in terms of functionality
then you have a virtual machine. This does not require using bytecodes
or an interpreter, such as java uses. The normal lisp incremental
compilation will work nicely.
CL does not provide this, and I suspect that scheme does not either,
but given care it should be possible to design a lisp system
so that it provides a virtual machine sufficient for transparantly portable
If you want to implement this system as an interpreter, fine, but just
remember that that's not necessary. The main reason for using JVM in java
is for portable binaries for transport.