[Python-Dev] Splitting up the PVM

[Guido van Rossum]

[me]
> > Here's what I think we should do first (I've mentioned this before but
> > nobody cheered me on :-).

[Jack]
> Go, Guido, GO!!!!
> 
> What I'd like in the split you propose is to see which of the items would be 
> implementable in Python, and try to do the split in such a way that such a 
> Python implementation isn't ruled out.

Indeed.  The importing code and the read-eval-print loop are obvious
candidates (in fact IDLE shows how the latter can be done today).  I'm
not sure if it makes sense to have a parser/compiler or the VM written
in Python, because of the expected slowdown (plus, the VM would
present a chicken-egg problem :-) although for certain purposes one
might want to do this.  An optimizing pass would certainly be a good
candidate.

> Am I correct in guessing that after factoring out the components you mention 
> the only things that aren't in a "replaceable component" are the builtin 
> objects, and a little runtime glue (malloc and such)?

I guess (although how much exactly will only become clear when it's
done).  I guess that things like thread-safety and GC policy are also
pervasive.

--Guido van Rossum (home page: http://www.python.org/~guido/)