Re: 3.99.3

From: Jay Carlson <nop_at_nop.com>
Date: Mon, 05 Oct 1998 15:45:29 -0500

> Everyone goes off into their own corner, assured that since we have
> proofs of concept for the wonderous compilers and interpreters, we
> need not actually bother creating useable ones for real world
> needs. Everyone who isn't an academic goes off and writes their own
> Perfect Scheme Interpreter, and thus we get twenty of them competing
> for users and implementation development. The result is, of course,
> that almost no one has heard of STk, very few have heard of Guile, and
> monstrosities like Perl continue to grow like fungii.

Let me state a popular perception, if not pure reality, in pithy form:

  CARLSON'S OBSERVATION

  There are more available Scheme implementations than available
  Scheme applications.

Let's rule out all the fine partial evaluators, compilers, inference
engines, and the zillions of not-really-interoperable object, record,
macro, and ADT library systems. Users don't see them, so they don't
count. (I refer to these kinds of packages as "incestuous", and Tcl
has a pile of them too.) Libraries for things like network services
and database access only sorta count.

The members of this mailing list know that Scheme is used (and useful)
for outside-of-Scheme (non-incestuous) research and development, but
the size of the audience for the resulting system may be small, or
even unity. This doesn't help build the public perception.
Meanwhile, CPAN and Tcl and Python are happy to point you to piles of
useful apps available for download.

Guile may be changing this. I'll remain skeptical until GNOME gets
userbase and stability.

> If the scheme community finally wants to win, it is going to have to
> accept the fact that inventing the perfect language is uninteresting
> if it has no users, and that it might be better to help your neighbor
> improve performance by a factor of two than to produce a fiftieth
> scheme interpreter to be placed on the scheme repository FTP site and
> forgotten.

Scheme will always be a little interesting even if the only users of
it are the theorists and developers of Scheme systems; nobody
complains that the pure lambda calculus doesn't have a good HTTP
server available, after all. It has demonstrated a useful place in
prototyping.

But: there are powerful benefits to having a large user base and a
perception of practicality and future, even to pure researchers.
Research related to languages in wide-spread use tends to be
better-funded.... :-)

Jay Carlson nop_at_nop.com
Received on Mon Oct 05 1998 - 22:46:18 CEST