Re: Translation for extension is a bad idea

["Greg J. Badros" <gjb at cs dot washington dot edu>]

Jonathan Bartlett <johnnyb@wolfram.com> writes:

<snip>

> >     Han-Wen> It's too bad that the space for window managers has
> >     Han-Wen> already been taken from GUILE, but maybe we can show the
> >     Han-Wen> power of GUILE with another application. That will
> >     Han-Wen> accomplish much more than any kind of translator will
> >     Han-Wen> ever do.
> > 
> > Actually, there's SCWM (you're aware of that, I assume).  Which makes
> > a bad point for Guile when compared to Sawfish (but that may or may
> > not be the fault of Guile).
> 
> Actually the author of Sawfish has mentioned going over to Guile.  He had
> a few problems with it (I can't remember them right now), but he said that
> if those were resolved, he would consider switching.

<snip>

As I see it, there are a couple of reasons why Sawfish has been more
popular than Scwm:

1) Performance (esp. startup) of Scwm is still poor.  I fault Guile
mostly here, as we all know that Guile's not got the fastest startup
time.  (Things are improved in 1.4, to be sure, but not enough, yet.)

2) Sawfish is cleaner inside (Scwm is based on fvwm2 and has a lot of
cruft -- we started with fvwm2 and converted, John started small and
built up, doing things right from the start... we were more
feature-complete earlier, but less agile moving forward).

3) Sawfish has better theme support (largely derives from point 2, above)

4) Sawfish benefits from John being the author of librep and being able
to exploit the extension language subsystem more completely.  Though
I've been (or at least once was) a substantial contributor to Guile, I
not only do I not understand Guile nearly as well as John understands
librep (actually, he understands librep probably better than anyone
understands guile), but I also have had to argue for the changes I
wanted to guile, rather than just make them myself.  I'm not complaining 
about that tradeoff (as I value the stability of Guile and recognize
Scwm is obviously and thankfully not the only client of guile), but just 
observing that it's a hindrance towards making improvements to Scwm that 
extend into the extension language.

5) Sawfish has intentionally stayed lean, rather than adding every new
cool module to the core distribution.

6) Some highly dynamic, extensible things are slightly more awkward in
Scheme than in Lisp.  (E.g., bindings that want to use symbols to bind
to are complicated by the module system; the alternative, using the
procedure object, means that when the procedure is re-defined, the
binding needs to be re-made.)

I'm curious what John's list of what he needs in Guile to want to switch 
Sawfish to using it is.  John? 

Thanks,
Greg
Scwm Maintainer
P.S. Expect a Scwm that works w/ Guile-1.4 to be released by month's end 
(already works in CVS, modulo some bugs w/ Fvwm2 modules)