Re: 3.99.3

From: Russ McManus <>
Date: 04 Oct 1998 13:44:15 -0400


First, let me say that I appreciate your comments Harvey, because I
know you from your many previous postings in scheme related forums, and
I've always found your comments insightful. (Harvey J. Stein) writes:

> As far as I'm concerned, both STk & Bigloo are substantially superior
> to Guile, in terms of speed, features & the quality of the underlying
> C implementation itself.

For the last year or so, I have mostly been using guile, and have
stopped using Bigloo and STk, so I am not a good judge of the relative
merits of these implementations in their current state. I take you at
your word on these issues.

> If your concern is embedding, it'd be better to add a convenent
> embedding mechanism to STk than to base STk on Guile.
> > Some of the work is already done; there is a significant piece of
> > STklos implemented and downloadable from
> >
> If your concern is the Tk interface for Guile, then you should just
> finish off the above.

This is very true; I could finish off the above work, and have a nice
Tk interface. But I am concerned about more than my immediate needs;
I would like to see the efforts of the STk maintainers and users reach
the widest possible audience. I think that an STk based on guile will
reach a wider audience than the current STk. And I don't just mean
today, I'm also taking into account the future (more on this below).

I acknowledge the fact that moving to guile would be a short term
loser. My assertion is that it would be a long term win to bring
together these two user communities. I think it can be accomplished
without too much pain.

> > Using guile has it's own set of issues, but there are some
> > considerable advantages, too. For one thing, there is a dedicated
> > maintainer for the interpreter itself. Also, there is a growing
> > developer community. There is a C compiler for Scheme code in guile
> > (not as nice as bigloo, I agree), too.
> This is misleading. Guile is still buggier than STk & Bigloo. Both
> STk & Bigloo also have growing developer communities. Both STk &
> Bigloo have dedicated maintainers.

I certainly didn't intend to mislead anyone; my regrets for that. I
did not mean to imply that guile is more stable than any other

I would like to refine one of my comments, though. I think guile has
a rate of growth significantly greater than STk. These things are
impossible to measure, of course. But in my admittedly subjective
view, I think that the growing number of important applications (some
GNOME applications come to mind) that are using guile as an extension
language is bringing along lots of new people.

I think that because guile is the GNU project's scheme implementation,
there will be many people who work on it and improve it, just because
of this. My assertion is that it is a good idea for STk to take
advantage of this future work.

> The Guile compiler (Hobbit) has lots of problems. For one thing, it
> compiles arithmetic to C artithmetic - not to scheme arithmetic. It's
> also a pain in the ass to use cleanly.

I'n no expert on hobbit. Off topic, but I guess that the math
compilation behavior could be coded both ways, and then turned into a
compiler switch.

You see, STk and Bigloo are both essentially the work of incredibly
smart and hard working _individuals_. Making their work compatible
with guile ensures that when these people move on to other interesting
stuff, all their work will continue to be maintained, albeit by
probably less bright, but still committed people.

kind regards,

"Don't worry about people stealing your ideas. If your ideas are any
good, you'll have to ram them down people's throats."
             -- Howard Aiken     
Received on Sun Oct 04 1998 - 19:45:37 CEST

This archive was generated by hypermail 2.3.0 : Mon Jul 21 2014 - 19:38:59 CEST