srini@skorpio3.usask.ca (P. Srinivas) writes:
> Is there any real reason other than egos that prevents all Schemer's
> rally behind Guile?
There are plenty of reasons to not rally behind Guile.
Most of us aren't strictly language implementors. We are also in the
business of being language _designers_. The point of using Scheme is
that we get a superb core language (superb for some purposes, anyway)
upon which we can build our languages of choice. Indeed, every major
Scheme system right now -- Chez, Gambit, Scheme 48, RScheme, MzScheme
and others -- each has some specific agenda items it is trying to get
a handle on.
> a) Everybody can implement her own version for
> testing various ideas. But at the same time all should
> rally for political reasons behind ONE implementation.
Yes, and we would each love nothing more than have everyone else rally
behind our own implementations!
It is not feasible to follow the approach you suggest. It would mean
having to duplicate all the effort on the "local" implementation as
well as the True one. If the two differ stylistically, all your test
programs would have to be written in both (or else how would you
conduct your experiments?). Et cetera. Life is too finite and too
interesting for that.
> c) When some features tested on owns privite scheme and works well,
> he should make sure that feature also works for the ONE TRUE
> Scheme. That way, we do not need to recomond some other Scheme than
> the ONE TRUE SCHEME for the sake of that perticular feature.
Assuming, that is, that the feature will work in that Scheme. For
instance, Scheme 48 has a very well developed module system. However,
when we designed MzScheme's system, we had some philosophical
disagreements with what Scheme 48 was doing. In the end, ours
overlapped for the most part, but not entirely. But now, who's to
tell that our system will even implementable on top of the True one?
> We can choose any existing Scheme for this purpose.
So says you. Apparently, the people doing the actual Scheme
implementation work don't quite agree.
> Is there any reason not to consider GUILE as the possible candidate
> for this?
Perhaps, instead, you could explain why we _should_ consider Guile:
what it offers that the other systems that are out there, actively
being developed, and are the targets of many, many hours of research
figuring out how to put them together right, don't offer. Beyond, of
course, the venerable names of the FSF and RMS.
There are Scheme systems that already offer foreign function
interfaces, macro systems, portability across numerous platforms,
platform-independent GUI libraries, embeddable interpreters, object
systems, module systems, type systems, programming environments,
static debuggers, program steppers ... with solid research behind
them, and often all in the same system. Why wait for Godot?
'shriram
|