Re: LFM + LFSP = LFE?

[Michael Vanier <address@hidden>]

Guile was originally intended to have multiple syntaxes so that, as you
mention, you could easily "map" other languages onto the scheme syntactic
constructs.  And as you say, this was too ambitious.  However, there was a
C-like dialect called CTAX which was not based on any other language
(though the syntax looked a lot like C).  This was actually implemented,
but it has totally bitrotted AFAIK.  I think they should have put more
effort into this and less into the language translation approach.  It would
have been much easier, and would probably have attracted many people to
guile.

Mike

> Date: Sat, 14 Jun 2003 18:47:46 -0400
> From: Adam Turoff <ziggy@panix.com>
> 
> On Fri, Jun 13, 2003 at 06:34:40PM -0700, Michael Vanier wrote:
> > Parenthetically (pun intended), the guile scheme people did this a few
> > years ago but abandoned the effort.
> 
> Which effort are you talking about?
> 
> I seem to remember a project that used Guile in conjunction with partial
> evaluation.  The idea was to map syntaxes like Perl, Tcl, Python, etc.
> onto scheme and have just the one runtime.  Nice in theory, but horrible
> in practice.
> 
> One reason why that project failed was that it was just too broad and
> ambitious: Perl folks are quite fond of pointing out that 'only Perl can
> parse perl'.  That is, the only way to interpret a Perl program is to
> implement all of the idiosyncratic behaviors the perl interpreter supports
> today.  Partial evaluation of Perl into Guile would involve faithfully
> reimplementing the perl, a holy grail that's been attempted multiple times
> but not yet achieved.
> 
> Z.
>