Re: [Sbcl-devel] SBCL roadmap

[William H. N. <wil...@ai...>]

On Tue, Jan 08, 2002 at 07:40:06PM -0800, JB wrote:
> I am following the SBCL project with interest, and I'm curious what's in
> store for the future.  What features/fixes/changes do you have yet to do,
> before you will be ready to call it "SBCL Version 1.0"?  I know there is no
> schedule here at all but in your own mind you must have a general timeframe
> for achieving that goal.  Or maybe not! :-), who knows.

Tentative ideas for things to do before 1.0 are in the TODO file in
the distribution.

The biggest single thing that I think needs to be done before 1.0 is
merging the CLOS code (in the SB-PCL package, built by the target
compiler in warm init) with the ordinary type system code (in the
SB-KERNEL package, built by the cross compiler). I'd be embarrassed to
call a Common Lisp system version 1.0 when it still had a totally
non-ANSI distinction between between SB-PCL:CLASS and CL:CLASS (and
similarly between SB-PCL:FIND-CLASS and CL:CLASS and a few other
symbols). And the natural way to fix this, I think, involves this
major restructuring of the way SBCL is built (no warm init any more),
of the way PCL is initialized (cross-compiling everything that needs
compilation for initialization), and very likely of the way that the
compiler's type system works also (using CLOS systematically instead
of using the twisty little hand-rolled pre-CLOS type method system).

Most of the other projects in TODO are of more manageable size. (There
are some other larger projects, e.g. speeding up the compiler, but
mostly they can at least be done in smaller steps of manageable size.)

I don't have even a general timeframe for this. Significant software
schedules always have some baseline uncertainty. When making big
changes to legacy software, the baseline is fairly high because you
can easily get blindsided by totally unexpected inflexibility or
unusually hard-to-fix bugs (as has happened a lot in 0.pre7). And
beyond the uncertainty in how many man hours it will take, I don't
know how much time either I or others will devote to it in any given
month. (I am pretty sure, however, that after 0.7.0, the amount of
time that I devote per month will drop for a while! In theory SBCL is
a means to an end to writing other applications, and they've been on
hold for much too long.)

-- 
William Harold Newman <wil...@ai...>
"Our users will know fear and cower before our software! Ship it! Ship
it and let them flee like the dogs they are!" -- Klingon programmer
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C  B9 25 FB EE E0 C3 E5 7C