Re: Emacs with Guile

[Per Bothner <per at bothner dot com>]

Per Bothner <per@bothner.com> wrote:
> Still, I do think the technical design of Kawa is much stronger than
> Guile.

That was an arrogant way of putting it (but that's me ...).
On my way home from a movie (Boys Don't Cry), that bothered me.
The key point is SCM made certain design decisions, focused
on being a very fast and clever interpreter.  My preference
is to implement a language as a compiler - but with fast response.

Mikael Djurfeldt <mdj@mdj.nada.kth.se> responded:

> Regarding Guile, I think it is improving.  There are a lot of people,
> like Shriram Krishnamurthi, who are complaining without actually
> knowing the state of art regarding Guile.  We are progressively
> attacking weak areas, and will, after some urgent priorities like
> manual and module system, commence a systematic revision of Guile's
> interfaces.

I absolutely agree.  The problem is that the core remains an
interpreter, which makes a lot of sense for a small fast extension
language.  Once you add a module system, classes, real macros, and so
on, you no longer have a small fast extension language, and a compiler
seems to make more sense.

By the way - I recently implemented a module system for Kawa.
(It's only in cvs and hasn't been documented yet.)
The simplest way to use it actually comes from Aubrey Jaffer:
Basically (provide MODULE) imports the names of MODULE into
the current scope, but does not re-export them.  This is
quite straight-forward to do with a compiler - the imported
names are managed at compile-time, without using run-time
"environments".
--
	--Per Bothner
per@bothner.com   http://www.bothner.com/~per/