Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Common Lisp Implementations: A Survey

20 views
Skip to first unread message

Daniel Weinreb

unread,
Dec 8, 2007, 9:05:36 AM12/8/07
to
There are ten currently-maintained implementation of Common Lisp.
What are they, and what are they like? What do they cost, are they
open-source, and where do you get them? What platforms do they run on?
Where did they come from? What platforms do they run on? Do they
support threads, streams, CLIM, the CLOS MOP, and/or profiling?
Answers to these questions and more can be found in my survey, at

http://common-lisp.net/~dlw/LispSurvey.html

There's also a glossary of terms, a table of some major Common Lisp
libraries showing which implementations are supported, a long list of
papers about Common Lisp, some success stories of Common Lisp
applications, a list of Common Lisp textbooks that are available
online, and an extensive list of Common Lisp resources, all with
links.

-- Dan Weinreb

Slobodan Blazeski

unread,
Dec 8, 2007, 10:07:16 AM12/8/07
to

Yesterdau I wrote about choosing the right lisp implementation
http://tourdelisp.blogspot.com/2007/12/choosing-right-lisp-implementation.html
though I was giving only a broad advice.
I will post an update including a link to your survey .

cheers
Slobodan

Javier

unread,
Dec 8, 2007, 10:15:26 AM12/8/07
to
Nice work, thank you.
It seems that actually SBCL is the best implementation.

Rainer Joswig

unread,
Dec 8, 2007, 11:54:30 AM12/8/07
to
In article <Qwx6j.2052$Bg7.1903@trndny07>,
Daniel Weinreb <d...@alum.mit.edu> wrote:

GCL is not deader than usual. It may smell funny.

Judging from the mailing list and from the CVS there
is still activity.

There is maintenance for GCL 2.6.8 and work
on GCL 2.7.0. GCL is still used by active projects like
Maxima and ACL2, AFAIK.

--
http://lispm.dyndns.org/

John Thingstad

unread,
Dec 8, 2007, 12:16:22 PM12/8/07
to

GCL and ECL seem to have a common root Kyoto Lisp.
What is the difference between them?

--------------
John Thingstad

Pillsy

unread,
Dec 8, 2007, 2:38:44 PM12/8/07
to
On Dec 8, 10:15 am, Javier <javu...@gmail.com> wrote:
> Nice work, thank you.
> It seems that actually SBCL is the best implementation.

I also really appreciate the resource.

However, I think the thing that stood out most was that there wasn't a
clear "best" implementation. As much as people complain about the
portability headaches that come from having a diversity of CL
implementations, and that certain desirable features aren't available
in a given combination, the list makes it obvious that there's also a
real variety of needs and priorities being served.

There's a large number of high-quality, actively developed, mature
Common Lisp implementations that are being actively developed. For
some reason this is regarded as a flaw in the language or in the
community, but I think it's a real strength, and one that's
particularly noteworthy given the relatively small size of the CL
community.

Cheers,
Pillsy

Klaus Schilling

unread,
Dec 8, 2007, 11:01:38 AM12/8/07
to
Javier <jav...@gmail.com> writes:


> It seems that actually SBCL is the best implementation.

no, it is a bad implementation, as it lacks a C API,
unlike CLisp and ECL.
It can't even be compiled from source with gcc and GNU tools.

In addition, it lacks support for first-class environments,
unlike Clisp.

Klaus Schilling

Cor Gest

unread,
Dec 8, 2007, 4:20:57 PM12/8/07
to
Some entity, AKA Pillsy <pill...@gmail.com>,
wrote this mindboggling stuff:
(selectively-snipped-or-not-p)

> However, I think the thing that stood out most was that there wasn't a
> clear "best" implementation.

Compare it like cars, a Ferrari is capable of transporting manure, but
a Rolls-Royce is realy better suited for is has a bigger booth.

Cor

--
Alle schraifvauden zijn opzettelijk, teneinde ieder lafaard de kans te
geven over spelling te zeuren in plaats van in te gaan op de inhoud.
Unless you are prepared to die, do not treathen me or my family
(defvar My-Computer '((OS . "GNU/Emacs") (IPL . "GNU/Linux")))
spam delenda est

Rupert Swarbrick

unread,
Dec 8, 2007, 4:48:00 PM12/8/07
to
> There is maintenance for GCL 2.6.8 and work on GCL 2.7.0. GCL is still
> used by active projects like Maxima and ACL2, AFAIK.

Not entirely sure, but I believe Maxima won't build with GCL at the
moment.

Rupert

Raymond Wiker

unread,
Dec 8, 2007, 6:00:48 PM12/8/07
to
Klaus Schilling <schilli...@web.de> writes:

> Javier <jav...@gmail.com> writes:
>
>
>> It seems that actually SBCL is the best implementation.
>
> no, it is a bad implementation, as it lacks a C API,
> unlike CLisp and ECL.

So, Clisp and ECL are better Common Lisp implementation than
SBCL because they have certain facilities/mechanisms that are not part
of Common Lisp?

> It can't even be compiled from source with gcc and GNU tools.

Oh, it can be compiled from source... you'll need a working
Lisp system (preferably SBCL, but certain other Lisps work, too). This
process is called bootstrapping, and you need certain infrastructure
before you have a self-hosting system.

> In addition, it lacks support for first-class environments,
> unlike Clisp.

So, Clisp is a better Common Lisp because it has certain
facilities/mechanisms that are not part of Common Lisp?

You're a fool, Klaus Schilling.

Pillsy

unread,
Dec 8, 2007, 6:45:04 PM12/8/07
to
On Dec 8, 6:00 pm, Raymond Wiker <r...@RawMBP.local> wrote:

> Klaus Schilling <schilling.kl...@web.de> writes:

> > Javier <javu...@gmail.com> writes:

> >> It seems that actually SBCL is the best implementation.

> > no, it is a bad implementation, as it lacks a C API,
> > unlike CLisp and ECL.

> So, Clisp and ECL are better Common Lisp implementation than
> SBCL because they have certain facilities/mechanisms that are not part
> of Common Lisp?

If you need those facilities to do what you want to do, or even if
it's just convenient for you, than yes. Every implementation on that
list provided features that go beyond the standard, including SBCL.
This suggests that people actually want those features.

Expecting people not to judge Common Lisp implementations on how they
differ is silly, and defeats the whole point of having multiple
implementations.

I'm going to go out on a limb and guess that Mr Schilling was pointing
that out in a slightly subtle way.

Cheers,
Pillsy

Daniel Weinreb

unread,
Dec 9, 2007, 9:55:35 AM12/9/07
to

Rainer,

Thank you for the extensive comments that you emailed
me today. I have made the revisions you suggested and
they will be in the next version of the survey, which
I'll post at the same URL.

More than one person has encouraged me to take another
look at GCL. I had simply taken a quick look at
www.gnu.org/software/gcl/ and saw that the last
"NEW!" announcement was dated Aug 10, 2005, which
seemed to be a statement that nothing had happened
in over two years. I also took a look at their
mailing list at http://lists.gnu.org/archive/html/gcl-devel/
and saw very little mail. It seems to end at version 2.6.7.

However, upon looking more closely, there is a little
traffic on the list. They are discussing 2.6.8 beta,
which I think means the development head in CVS. Camm
Maguire, whom I gather is the primary (only?) developer,
is anwering questions, and sometimes saying things that
seem to indicate that he's working on the implementation.
I'll send him email.

-- Dan

Rainer Joswig

unread,
Dec 9, 2007, 11:29:14 AM12/9/07
to
In article <HlT6j.1243$sf.1023@trndny04>,
Daniel Weinreb <d...@alum.mit.edu> wrote:

Btw., you can see many Lisp mailing lists here in one
place:

http://dir.gmane.org/index.php?prefix=gmane.lisp

I usually read those via a news reader, since gmane
is available also by NNTP.

Here you see GCL commits:
http://dir.gmane.org/gmane.lisp.gcl.cvs
This is the GCL mailing list:
http://dir.gmane.org/gmane.lisp.gcl.devel

It's getting less traffic. Though I think GCL is
an interesting implementation.

>
> However, upon looking more closely, there is a little
> traffic on the list. They are discussing 2.6.8 beta,
> which I think means the development head in CVS. Camm
> Maguire, whom I gather is the primary (only?) developer,
> is anwering questions, and sometimes saying things that
> seem to indicate that he's working on the implementation.
> I'll send him email.
>
> -- Dan

--
http://lispm.dyndns.org/

Pillsy

unread,
Dec 9, 2007, 3:50:13 PM12/9/07
to
On Dec 9, 11:29 am, Rainer Joswig <jos...@lisp.de> wrote:
[...]

> It's getting less traffic. Though I think GCL is
> an interesting implementation.

What does GCL offer that ECL doesn't? AIUI, in terms of both pedigree
and features, they're very similar.

Cheers,
Pillsy

Andy Chambers

unread,
Dec 9, 2007, 4:37:05 PM12/9/07
to
On Dec 8, 4:01 pm, Klaus Schilling <schilling.kl...@web.de> wrote:

> Javier <javu...@gmail.com> writes:
> > It seems that actually SBCL is the best implementation.
>
> no, it is a bad implementation, as it lacks a C API,
> unlike CLisp and ECL.
> It can't even be compiled from source with gcc and GNU tools.

Maybe I'm just a rubbish coder but I prefer to have a little more
information in my backtraces than clisp provides.

Andy

Maciej Katafiasz

unread,
Dec 9, 2007, 5:50:47 PM12/9/07
to
Den Sat, 08 Dec 2007 14:05:36 +0000 skrev Daniel Weinreb:

> There are ten currently-maintained implementation of Common Lisp. What
> are they, and what are they like? What do they cost, are they
> open-source, and where do you get them? What platforms do they run on?
> Where did they come from? What platforms do they run on? Do they
> support threads, streams, CLIM, the CLOS MOP, and/or profiling? Answers
> to these questions and more can be found in my survey, at
>
> http://common-lisp.net/~dlw/LispSurvey.html

The GCL thing has been already pointed out by Rainer, so I'll only
mention other things:

- You don't set <title></title>, which makes the window title presented
by the browser to be less than helpful.

- Not really a correction, but I just want to make sure I read correctly:
In CMUCL maintainers, you list Bruno Haible and Michael Stoll, whom I
instantly recognised to be Clisp authors. Is it correct?
... except that now I read it, it doesn't say that anymore, I guess this
was one of those private email corrections by Rainer?

Cheers,
Maciej

Waldek Hebisch

unread,
Dec 10, 2007, 7:54:49 AM12/10/07
to

1) Dumping running Lisp to an image
2) Loading compiled Lisp without using dlopen (at least some
implementations of dlopen could run of file descriptors
when loading many file)

--
Waldek Hebisch
heb...@math.uni.wroc.pl

Raymond Toy (RT/EUS)

unread,
Dec 10, 2007, 10:05:30 AM12/10/07
to
>>>>> "Rupert" == Rupert Swarbrick <rswar...@gmail.com> writes:

>> There is maintenance for GCL 2.6.8 and work on GCL 2.7.0. GCL is still
>> used by active projects like Maxima and ACL2, AFAIK.

Rupert> Not entirely sure, but I believe Maxima won't build with GCL at the
Rupert> moment.

AFAIK, it's still the default compiler for Maxima, and Maxima did
build for me with gcl a short time ago. But I have a rather old gcl.

Ray

Juanjo

unread,
Dec 10, 2007, 1:36:23 PM12/10/07
to
On Dec 10, 1:54 pm, Waldek Hebisch <hebi...@math.uni.wroc.pl> wrote:
> 1) Dumping running Lisp to an image
> 2) Loading compiled Lisp without using dlopen (at least some
> implementations of dlopen could run of file descriptors
> when loading many file)

The thing is _nothing_ prevents you from installing such features in
ECL. The reason why GCL has them is because it has implemented its own
loader of object files, with dynamic linking etc. This is highly
nonportable, but as a feature, it is a rather self-contained code
which can be ported to ECL.

IIRC, there is even a GNU library that implements such a thing.
However, as a maintainer of this implementation, I would rather stay
away from such a code, that forces one to follow closely the guts and
developments of all supported operating systems.

Juanjo

Daniel Weinreb

unread,
Dec 10, 2007, 11:56:19 PM12/10/07
to
Maciej Katafiasz wrote:
> Den Sat, 08 Dec 2007 14:05:36 +0000 skrev Daniel Weinreb:
>
>> There are ten currently-maintained implementation of Common Lisp. What
>> are they, and what are they like? What do they cost, are they
>> open-source, and where do you get them? What platforms do they run on?
>> Where did they come from? What platforms do they run on? Do they
>> support threads, streams, CLIM, the CLOS MOP, and/or profiling? Answers
>> to these questions and more can be found in my survey, at
>>
>> http://common-lisp.net/~dlw/LispSurvey.html
>
> The GCL thing has been already pointed out by Rainer, so I'll only
> mention other things:
>
> - You don't set <title></title>, which makes the window title presented
> by the browser to be less than helpful.

Thanks, I'll do this.

>
> - Not really a correction, but I just want to make sure I read correctly:
> In CMUCL maintainers, you list Bruno Haible and Michael Stoll, whom I
> instantly recognised to be Clisp authors. Is it correct?
> ... except that now I read it, it doesn't say that anymore, I guess this
> was one of those private email corrections by Rainer?

Yes, someone already reported that and I fixed it. Thanks.

>
> Cheers,
> Maciej

Camm Maguire

unread,
Dec 12, 2007, 12:17:13 PM12/12/07
to
Greetings!

Rupert Swarbrick <rswar...@gmail.com> writes:

Not to my knowledge.

Take care,

> Rupert

--
Camm Maguire ca...@enhanced.com
==========================================================================
"The earth is but one country, and mankind its citizens." -- Baha'u'llah

Camm Maguire

unread,
Dec 12, 2007, 12:22:47 PM12/12/07
to
Greetings!

"John Thingstad" <jpt...@online.no> writes:

Biggest difference is that GCL has tried to keep the traditional LISP
repl model, for example supporting the runtime load of executable
objects directly into managed memory, which can then be saved to disk
and restarted later. GCL effectively carries around its own native
object linker for this purpose. ECL is better at embedding lisp into
a C programming environment.

Take care,

> --------------
> John Thingstad

Camm Maguire

unread,
Dec 12, 2007, 12:24:30 PM12/12/07
to
Greetings!

Daniel Weinreb <d...@alum.mit.edu> writes:

> Rainer Joswig wrote:
> > In article <Qwx6j.2052$Bg7.1903@trndny07>,
> > Daniel Weinreb <d...@alum.mit.edu> wrote:
> >
> >> There are ten currently-maintained implementation of Common Lisp.
> >> What are they, and what are they like? What do they cost, are they
> >> open-source, and where do you get them? What platforms do they run on?
> >> Where did they come from? What platforms do they run on? Do they
> >> support threads, streams, CLIM, the CLOS MOP, and/or profiling?
> >> Answers to these questions and more can be found in my survey, at
> >>
> >> http://common-lisp.net/~dlw/LispSurvey.html
> >>
> >> There's also a glossary of terms, a table of some major Common Lisp
> >> libraries showing which implementations are supported, a long list of
> >> papers about Common Lisp, some success stories of Common Lisp
> >> applications, a list of Common Lisp textbooks that are available
> >> online, and an extensive list of Common Lisp resources, all with
> >> links.
> >>
> >> -- Dan Weinreb
> > GCL is not deader than usual. It may smell funny.
> > Judging from the mailing list and from the CVS there
> > is still activity.
> > There is maintenance for GCL 2.6.8 and work
> > on GCL 2.7.0. GCL is still used by active projects like
> > Maxima and ACL2, AFAIK.
> >

Yes, it is my intention to finalize 2.7.0 reasonably soon. I had
thought this fall, but other priorities have intervened.

Take care,

0 new messages