OFF-TOPIC:: Why Lisp is not my favorite programming language

nobody nobody_u_should_no at yahoo.com
Wed Mar 3 13:43:03 CET 2004


This article is posted at the request of C.W. Yang who
asked me to detail my opinion of Lisp, and for the benefit
of people like him, who may find themselves intrigued by
this language.

The opinions expressed herein are my personal ones, coming
from several years of experience with Lisp. I did plenty of
AI programming back in the day,  which is what would now be
called "search" instead.

Due to time constraints, I will refrain from posting any
follow-ups here. Participants of other newsgroups are well
aware of comp.lang.lispers' tendency to engage in personal
attacks, so a productive discussion with them is unlikely
anyway.

Permission is granted to copy and distribute this document
without restrictions.


=========================================================
   Why Lisp is not my favorite programming language.
=========================================================
(In the following, "Lisp" refers to ANSI Common Lisp)

This article is a collection of facts anyone interested in
Lisp should know about.

---------------------------------------------------------

FACT 1: The fastest Lisp implementations are slow

(See any third-party benchmark: The Great Computer Language
Shootout comes to mind, but the Coyote Culch test is in my
optinion even better: it is a professional-quality
interlanguage benchmark)

As a rule of thumb, the most hand-optimized Lisp programs
will be longer than their C/C++ equivalents, and will run
2-20 times slower using the best compilers.

This alone is half the truth. To get such performance out
of Lisp, one has to add type declarations and shed all
safety checks, which is analogous to casting /everything/
to (void*) in C. This is needed to turn off type tag checks
at run time.

Sadly, Lisp code that was posted to USENET by Pascal
Bourguignon for the Coyote Gulch test did not measure up
even to these low expectations and was 31.6 times slower
than C++, using CMUCL 18d as a Lisp compiler and Intel C++
7.1 (with -O3 -xW optimization switches) as a C++ compiler
on Pentium IV.

---------------------------------------------------------

FACT 2: No one but a small clique of fanatics likes it

No matter how odd or perverted the cause, there will be
followers. In fact, the odder the cause, the fewer, but
more fervent its followers are. Look at any religious cult,
like Scientology, or take a peek in comp.lang.lisp and
listen to Lisp zealots talk about 'making sacrifices for
the cause' (in all seriousness!). Look up "a public
apology" thread started by Pascal Costanza as an excellent
example.

ZEALOTS STOP AT NOTHING TO ADVANCE THEIR CAUSE, AND
PERSONALLY ATTACKING ANYONE WITH A DISSENTING OPTINION IS
THE FIRST THING IN THEIR ARSENAL.

---------------------------------------------------------

FACT 3: The vast majority of people who study Lisp in
school, never want to use it again.

You should already know this if you studied or taught CS
where Lisp courses were offered. Even those students who
are fond of Scheme are usually disgusted by Lisp.

---------------------------------------------------------

FACT 4: Lisp is the most complicated language in the world

It has the biggest standard specification document, which
is also the most obfuscated one - something a lawyer
pretending to be a programmer could have written. C, C++
and Fortran 95 specs are much better written, by people who
can communicate directly and eloquently.

---------------------------------------------------------

FACT 5: Despite its size, Lisp does not define threads or
GUI.

Large libraries are very useful when programming, however
Lisp's many functions and macros hardly qualify.

---------------------------------------------------------

FACT 6: There is no open-source cross-platform native-code
compiler

It was suggested that GCL (GNU Common Lisp) is the only
exception. However, it needs to be noted, that despite its
name, GCL is a dialect of its own, is quite slow even by
Lisp standards, and most alarmingly, unlike with other
compilers, its license requires your, programmer, code to
be GPL if you distribute it with GCL. (Because you will
need to use GCL both as a compiler and a run time library)

---------------------------------------------------------

FACT 7: There is no standard C interface.

C has become a lingua franca for interlanguage APIs. It may
be unfair, but not having a standard C interface is a
serious problem for any language.


EPILOGUE:

I do not hate Lisp, and I think it was a fine tool decades
ago, and I am not going to say "Lisp sucks". However, now
that we have superior languages for coding close to the
iron, high-performance computing, number crunching,
algorithms, scripting and gluing components together, Lisp
should be finally retired.



More information about the Python-list mailing list