> Date: Mon, 2 Oct 1995 05:53:44 -0400
> Reply-To: [EMAIL PROTECTED]
> From: Manuel Chakravarty <[EMAIL PROTECTED]>
>
> > To me, one of the most regrettable characteristics of
> > the Algolic family of languages is the tendency of the
> > compiler to turn into a giant black box of facilities
> > open only to an elite minority of compiler hackers, which
> > then begins inexorably sucking the entire programming
> > support environment down its event horizon.
> >
> > I would much prefer that the concept of "compiler" in this
> > sense did not exist, and that instead one had a nicely
> > factored translation toolset wide open to the application
> > programmer. Lisp and Forth begin to approach this ideal.
>
> Would you mind divulging the identity of your hilarious correspondent?
I got the impression that his original mail was distributed to the whole
Haskell mailing list. Anyway, I append it at this message.
Cheers,
Manuel
P.S.: As it seems that there are a number of people who didn't get the mail I
responded to, I CC this to the whole mailing list. Sorry, for any
duplicates.
-------------------------------------------------------------------------------
Date: Sat, 30 Sep 1995 10:09:08 -0700
From: [EMAIL PROTECTED] (Jeff Prothero)
To: [EMAIL PROTECTED]
Subject: Re: Haskell 1.3: modules & module categories
Cc: [EMAIL PROTECTED]
Manuel Chakravarty <[EMAIL PROTECTED]> writes:
| [...] it is desirable to be able to
| restrict the access to some modules in a way that the
| compiler can control when a group of people is working
| in one module hierarchy. Too illustrate this, assume
| that we classify the modules into different levels of
| abstraction, say, three levels: [...]
To me, one of the most regrettable characteristics of
the Algolic family of languages is the tendency of the
compiler to turn into a giant black box of facilities
open only to an elite minority of compiler hackers, which
then begins inexorably sucking the entire programming
support environment down its event horizon.
I would much prefer that the concept of "compiler" in this
sense did not exist, and that instead one had a nicely
factored translation toolset wide open to the application
programmer. Lisp and Forth begin to approach this ideal.
At the least, it would be very nice if the compiler
could be kept distinct enough from the rest of the
programming support environment that it doesn't begin
sucking what sound to me like logically separate
project management concerns (above) into its orbit.
Would it be possible to define an interface which
allows the above sort of "Not if you're a left-handed
programmer and it's Tuesday" restrictions to be
separately implemented and kept out of the core
language?
(To my mind, one of the successes of C -- as distinct from
C++, say -- is that it clearly defined what was and wasn't
the task of the compiler, and stuck to its guns, resulting
in that very rare bird: An Algolic language with a stable
language definition and compiler.)