Re: About modules

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

Lalo Martins <lalo@hackandroll.org> writes:

> On Tue, Mar 07, 2000 at 12:38:59AM +0100, Mikael Djurfeldt wrote:
> > Lalo Martins <lalo@hackandroll.org> writes:
> > 
> > > Nice to know Guile has officially been declared "low priority".
> > 
> > It hasn't.
> 
> Sorry, I was only trying to provoke some reaction. I mean, it's
> a quite big project... if Maciej is busy, I'm sure there is
> someone else with the power to merge such an important patch?

I know Mikael is capable of doing this merge: Mikael, do you have the
time?  If not, and if we're all agreed that the patch should go in ASAP,
I'll do it (I really don't have the time, but I'll do it anyway as long
as others will be able to help with the inevitable fallout).

> Guile is a moving target currently. It isn't worth coding for
> it, because environments, goops and the new module system will
> change the world, and render all existing code not
> incompatible, probably, but obsolete.

Well, the guile devs certainly care about the existing base of code --
hell I've got 23,000 lines of Guile to maintain in Scwm! 

> Still, the changes just don't happen. Which means, people
> choose not to code for Guile, because they can't wait.

This is a big problem, for sure.  

> For example, if Guile was reasonably usable 6 months ago,
> Sawmill would be based on Guile instead of Rep. Not to say SCWM
> isn't cool, but Sawmill quickly climbed to the top-5 list,

Out of curious, which list? 

It's definitely clear that Sawmill has gained popularity surprisingly
(and annoyingly, really! :-) ) quickly.  I attribute this in large part
to its minimalist design and early focus on usability -- John has done a
terrific job in keeping the distribution clean and simple.  He also
benefits from being intimately aware of the details of his interpreter
--- Guile, in part due to the lack of documentation, has been a steeper
learning curve for me and made using it tougher (especially in the early
months after Maciej's departure from Scwm, before I grokked Guile as
well as I now do).

Interestingly, some of the features that make Guile nicer as a
programming language complicate its use as an extension language (e.g.,
modules can get in the way and be confusing to use in conjunction with
the Scwm preferences module).

> _because_ it's so customizable (and SCWM is more, which is hard
> to understand). And a lot of projects using Python would be
> using Guile instead, too.

I agree.

We need to get the infrastructure for the major changes to Guile in
place to avoid the FUD (fear uncertainty and doubt) that is causing
extension authors to go elsewhere.  The environment patches, GOOPS, some
technique for speeding up loading and initialization of modules, and the 
documentation overhaul are the most important issues IMO.  

Incidentally, we've made some really good progress on adding docstrings,
but I need the help of the some of the Guile authors to document the
many not-standard-Scheme functions that are still missing
docstrings. (E.g., stacks.c and debug.c would probably only take 30
minutes each for the original author [Mikael, I believe] to add
docstrings to, but would require an hour and a half of my time, and I'd
make more superficial remarks and not necessarily have the perspective
to remove code that's no longer relevant.)

Greg