On Fri, Dec 20, 2002 at 03:52:41PM -0600, William Harold Newman wrote:
> On Fri, Dec 20, 2002 at 08:01:03PM +0000, Daniel Barlow wrote:
> > The NEWS file doesn't list a lot more stuff to happen in 0.7.x.
> > Anyone have any particular plans, or is it time to start thinking
> > about calling an upcoming release 0.8? Over the next few months I'd
> > like to do (a) threads, and (b) gencgc (probably without the c, though)
> > on non-x86 platforms.
> Do you mean NEWS or TODO?
I suspect TODO, though there is that stuff at the bottom of NEWS which
he could be referring to...
My take on what's left to do, for what it's worth, is that
0.7.0-0.8.0 involves essentially compiler infrastructure fixes.
0.8.0-0.9.0 involves modernization of PCL (which will, I suspect,
involve rationalizing its build system in much the same way as cmucl's
build system was rationalized, as well as working on many of the aspects
that Gerd has been dragging into the 1990s :-). I have ideas for the
clean bootstrapping of PCL, which I think would require only a
conforming CLOS (i.e. not a MOP) in the host compiler, but they're
currently sketchy (they were written down during a lecture about the
behaviour of the Radion -- a scalar field in 5-dimensional cosmology
0.9.0-1.0.0 would then involve
find . -name '*.lisp' | xargs grep FIXME
and dealing with the myriad of things that that reveals. :-/
On the shorter term, there are various things out there that I wouldn't
like to release 0.8.0 without:
* preliminary NetBSD support (patch from Valtteri Vuorikoski);
* working MAKE-INSTANCE (probably involving wholesale replacement of
fast-init.lisp with Gerd Moellmann's ctor.lisp);
* preliminary XREF support (from Eric Marsden; this addresses one of
the items on TODO -- "automatic mechanisms for finding dead code" as
well as introducing the possibility of supporting Naggum-desired
DEFMACRO behaviour :-);
* DEBUG-RETURN, which I played with this afternoon and which is _so_
_cool_ (from Frederik Kuivinen);
to which I would add, presuming that Dan's fast rate of progress
* preliminary thread support (creation and locking primitives, no
guarantees about thread safety, maybe as a build-time option).
Obviously, the broad-brushstrokes plan ignores various ideas on the
horizon, such as: a real 64-bit port (maybe on the alpha, maybe on a less
moribund platform); support for more than 128 (or 256 if you're lucky)
characters; improvements to the runtime (so that we can save cores with
foreign references -- our CMUCL friends have done some work in this area
:-); and hunting down performance bottlenecks and exterminating them,
along with doing work to support various other platforms that we might
want to support (MacOS, Windows, even -- possibly -- HP-UX or IRIX)?
Now, the good thing about a monthly release schedule is that it doesn't
matter too much if the "short term" plans take another couple of months
(though ideally not much more than that); we still release often enough
to get relatively frequent feedback and to remain at the back, if not
the front, of lisp users' minds.
Does this seem reasonable? Megalomaniacal? Unambitious? Have I
forgotten anyone or any exciting new development?
 worth remembering that my administrative privileges date from
/after/ the release of sbcl-0.7.0, at the very least, and so might not
jibe with others' expressed before then (or indeed since).
http://www-jcsu.jesus.cam.ac.uk/~csr21/ +44 1223 510 299/+44 7729 383 757
(set-pprint-dispatch 'number (lambda (s o) (declare (special b)) (format s b)))
(defvar b "~&Just another Lisp hacker~%") (pprint #36rJesusCollegeCambridge)