Re: Emacs rewrite in a maintainable language

[Stephen J. Turnbull]

Ken Raeburn writes:

 > Do you think the introduction of lexical binding support changes
 > the balance much?

FWIW, I was thinking that same thing but decided I am sufficiently
inexpert that I'd add more noise than signal.  Now that Ken's posted,
I'll stick my neck out and propose an empirical definition of RELisp
(that's not "Lisp with two problems", that's "restricted Emacs Lisp").

Specifically, have a Lisp-to-C (or even gcl-to-Evers-to-Chance-to-C,
as somebody-whose-name-I-should-remember-mea-maxima-culpa suggested)
compiler that just barfs on any non-lexical reference (and on defvar).

Learning it would be really easy: Just learn Emacs Lisp and then by
trial and error learn how to placate the compiler. :-)

Of course, I suppose that lexical refs probably already use a different
bytecode and an optimized code path, so compiling pure lexical code to
C would buy only a small integer factor of speedup at best.  But that
could be enough to make some mildly annoying delays fall within human
JND.

And the idea could easily be weakened to a bytecompiler mode that
refuses to emit dynamic refs, to teach people to avoid dynrefs on pain
of a cranky compiler. ;-)

Steve