[comp.sys.next] Allegro Common Lisp vs. Kyoto Common Lisp

tenny@ootool.dec.com (Dave Tenny) (06/20/91)

In article <1991Jun19.121755.27992@engage.pko.dec.com> tenny@ootool.dec.com  
I wrote:
[... stuff deleted ...]
> I may eventually shell out for the ACL 4.0 release. But it isn't what I need
> now.  All I need is my basic 3.1 CLTL Common Lisp.  And no, Kyoto Common Lisp
> is not an acceptable substitute.

To which eckel@ircam.fr (Gerhard Eckel) replied with the question:
[... stuff deleted ...] 
Just out of curiosity, why is KCL not an acceptable substitute? I never worked  
seriously with any (ACL & KCL) of them but I might in the future. Could  
somebody point out the differences for me?
-----------------------------------------------------------

Here is my reply, as a <large-computer-manufacturer>-Lisp developer of 
many seasons (but no longer), and also as a believer that Common Lisp
(with CLOS) is the BEST language for large scale problem solving available 
to date.

If you intend to use Lisp for solving real problems, then you require
a robust, efficient and well supported language for the job.  There is no
substitute for a finely honed architecture-specific lisp implementation,
unless you have mips, time, and memory to spare.

KCL is not particularly efficient as Common Lisp implementations go.
The price is right (free), and the portability is nice.  But it is a classic
case of getting what you pay for.  I can't speak for the truth of that statement
for the commercial lisps which have since been based on the original
KCL.

By way of analogy, using KCL is to Lisp as using "Ralph's Objective C" 
is to Objective C on the NeXT with NeXTstep.  A supported, well integreted,
well tuned implementation is much preferable.

Naturally, my opinion is worth what you paid for it as well, but I hope
it helps answer your question.  If you just want to write 
relatively inconsequential toy code in Lisp,
then KCL will probably meet your needs admirably.

Dave

[Lisp, the One, True Language!]

As an aside, people who favor Objective-C as being a much nicer language
to use than C++ will no doubt agree that part of the reason it is nicer
is the lack of rediculously strict typing that only language lawyers can
use in C++.  Implicit casting from id to XXX and back in Objective-C
and the runtime information available about classes is much more lisp-like
than anything C++ will ever have.

One more biased statement, since I'm on a Lisp soapbox today,
SmallTalk is just an amputated Lisp!

:-)

verket@venice.SEDD.TRW.COM (Paul Verket) (06/24/91)

In article <1991Jun20.141011.21744@engage.pko.dec.com> tenny@ootool.dec.com (Dave Tenny) writes:
...
>If you intend to use Lisp for solving real problems, then you require
>a robust, efficient and well supported language for the job.  There is no
>substitute for a finely honed architecture-specific lisp implementation,
>unless you have mips, time, and memory to spare.
>
>KCL is not particularly efficient as Common Lisp implementations go.
...
I moved some code from AKCL on a Sun 3 to Allegro on an 030 NeXT. An
informal timing test showed and 8 fold speed increase. Your mileage
WILL vary.

When I had a problem with Allegro's C interface, Franz Inc. debugged it in
a couple of days. In addition, their upgrade policy has been _very_
affordable for me as an individual.

Paul Verket 
verket@venice.sedd.trw.com (NeXTmail)