Why, Indeed?

Byron Davies davies@pobox.com
Wed, 25 Mar 1998 19:51:10 -0700

Let's go a little deeper into the "Why LispOS?" question.  As happened a
year ago, the technical discussion about LispOS is not leading anywhere in
particular.  I think the Common Lisp/Scheme split is a little nastier this
time around, but otherwise it's the same-old, same-old.  There seems no
more hope of reconciliation this year than last, so the best thing to do is
to create a second list, SchemeOS, and separate the two sides.  Kent
Pitman, a key contributor to the parenthesized world, says he ignores all
mail addressed to both comp.lang.lisp and comp.lang.scheme.  Surely, he
ignores this list too.  At least with separate lists, each crowd can
concentrate on technical issues, and perhaps generate more light than neat.

Unfortunately, I don't think that separation will move us any closer to a
new OS.  Some will ask, if Linux can evolve into a serious Unix, then why
can't LispOS (or SchemeOS) happen?

First, no driving need.  The driving need behind Linux was for a low-cost
Unix -- Unix has become the standard OS for non-trivial software
development, but the marketing strategies of Sun, HP, IBM, etc. didn't
allow for a sub-$100 Unix.

Second, no driving force.  The driving force behind Linux was the
determination of first Richard Stallman and then Linus Torvald to provide
free competition for the proprietary Unix vendors.

Third, demographics.  There just aren't enough Lisp afficianados with time
on their hands to make significant progress.  Counterexamples, of course,
may be found in the CMUCL and RScheme efforts, but their rates of progress
aren't sufficient to overtake PERL much less C++ or Java.  Furthermore,
unless coopted to one of the LispOS/SchemeOS efforts, those groups simply
serve to divide up the limited resources.

How did the original LispOS, the MIT Lisp Machine system, arise?  The
driving need for the MIT LispM was for a larger address space for
Lisp-based AI applications that were rapidly outgrowing the PDP-10's 18
bits of addressing.  The driving forces behind the MIT LispM were a team of
programmers with vision and extraordinary ability, and the financial
resources of DARPA.  The DARPA support began with the direct support of
LispM development, and continued for nearly a decade by providing AI
companies and AI labs within other companies with the money to buy
commercial LispMs.

What could drive a LispOS?  Although I'm a Lisp believer -- and a 25-year
Lisp user, including a heady decade with LispMs --  I don't expect to see
an outcry for a Lisp-based OS, or even a freeware Lisp environment.  Lisp
has about as much chance in the programming language world as the Mac has
in the  desktop computer world, for much the same reasons: (1) it has an
extreme market share disadvantage, and (2) it hasn't kept up with external
advances in software technology.  Lisp still has advantages over other
programming languages, but they don't make up for the disadvantages -- for
most people.  There will always be Lisp users (and Mac users) but most of
the world won't care.

Something could still happen, however.  For example, the leading vendor of
enterprise management systems is SAP, about 25% as big as Microsoft in
revenue.  In order to develop their industry-leading product, R/3, they
developed a CUSTOM 4GL which is only used within SAP, by about 2000
programmers.  This custom programming language is one of the reasons they
were able to progress so rapidly to market leadership.

Some had hopes, with CL-HTTP, that Lisp would play a similar role for
web-servers, but it didn't happen.  For lack of marketing by MIT, no
commercial entity was willing to make the investment necessary to convert
CL-HTTP into a whole product.

Where is the killer app that could make Lisp a winner again?  Rather than
aimlessly waving the technology wand, the Lisp community needs to search
for markets that will sustain the ongoing effort needed to bring Lisp up to
where it needs to be and keep it there.