Steel Bank Common Lisp

mikel <mikel@...> writes:
...
> Do you folks, the CMUCL and SBCL implementors, have any advice or 
> information that you think might influence the decision that the Skate 
> group makes about what Lisp to use for implementation?

Could you perhaps clarify why you need to choose a single
implementation?  Maybe that would make it obvious which would be best.

-- 
-Tim

Tim Daly Jr. wrote:

> mikel <mikel@...> writes:
> ...
> 
>>Do you folks, the CMUCL and SBCL implementors, have any advice or 
>>information that you think might influence the decision that the Skate 
>>group makes about what Lisp to use for implementation?
> 
> 
> Could you perhaps clarify why you need to choose a single
> implementation?  Maybe that would make it obvious which would be best.
> 

Each candidate Lisp is missing some required features, and the missing 
features vary. The more Lisps we support, therefore, the more missing 
features we must implement.

Therefore, to start with, we are trying to choose a single reference Lisp.

It is our express intention to make Skate as portable as possible, 
across Lisps as well as operating systems, and there has already been 
some preliminary discussion of what would need to be in an abstraction 
layer to facilitate portability. However, early and frequent release is 
a goal, so it behooves us to choose a Lisp the is as close as possible 
to ideal.

Again, the big missing features seem to be Unicode support and Win32 
support. SBCL has the first, and CMUCL seems about to have the second. 
Any other thoughts anyone cares to share would be welcome.

--me

Mikel <mikel@...> wrote:

> I just created an open-source project called Skate.

I have two questions about this (and no answer for the choice of
implementation, sorry):

- Is there a mailing list or something where interested lurkers like
  me can follow the discussion?  If not, would you like to create one
  (maybe on common-lisp.net)?

- If I understand you correctly, you don't want to build Skate on top
  of CLIM.  I understand that the higher levels of CLIM may be different
  enough from Skate that it makes no sense to use them, but is there a
  reason for not using at least the low-level CLIM stuff?  And if you're
  not going to use CLIM for this, what _are_ you going to use?

Thanks,

Arthur Lemmens

Mikel <mikel@...> wrote:

> Again, the big missing features seem to be Unicode support and
> Win32 support.

If "Win32 support" means "runs on Win32 systems", then I think CLisp
has both Unicode and Win32 support.  If "Win32 support" means "a usable
interface to a reasonable subset of the Win32 API", my impression is
that no open source Lisp has this.  (But I've never used CLisp or SBCL
or CMUCL, so please ignore me if I'm talking nonsense.)

Arthur

From: Arthur Lemmens <alemmens@...>
   Date: Mon, 15 Nov 2004 21:12:56 +0100


   Mikel <mikel@...> wrote:

   > Again, the big missing features seem to be Unicode support and
   > Win32 support.

   If "Win32 support" means "runs on Win32 systems", then I think CLisp
   has both Unicode and Win32 support.  If "Win32 support" means "a usable
   interface to a reasonable subset of the Win32 API", my impression is
   that no open source Lisp has this.  (But I've never used CLisp or SBCL
   or CMUCL, so please ignore me if I'm talking nonsense.)

CMUCL and SBCL do not run on Windows at all, as far as I know.

Could you detail which open source Lisp implementations you considered,
and could you say specifically what you mean by "usable interface"?

Arthur Lemmens writes:
> Mikel <mikel@...> wrote:
> 
> > Again, the big missing features seem to be Unicode support and
> > Win32 support.
> 
> If "Win32 support" means "runs on Win32 systems", then I think CLisp
> has both Unicode and Win32 support.  If "Win32 support" means "a usable
> interface to a reasonable subset of the Win32 API", my impression is
> that no open source Lisp has this.  (But I've never used CLisp or SBCL
> or CMUCL, so please ignore me if I'm talking nonsense.)

I don't use MS-Windows, but clisp has a complete FFI that should allow
you to access the Win32 API.

-- 
__Pascal Bourguignon__                     http://www.informatimago.com/
The world will now reboot; don't bother saving your artefacts.

On Nov 15, 2004, at 12:04 PM, Arthur Lemmens wrote:

> Mikel <mikel@...> wrote:
>
>> I just created an open-source project called Skate.
>
> I have two questions about this (and no answer for the choice of
> implementation, sorry):
>
> - Is there a mailing list or something where interested lurkers like
>  me can follow the discussion?  If not, would you like to create one
>  (maybe on common-lisp.net)?

Yes and no. There is not yet a real mailing list; just an alias I'm 
using. Peter Seibel says he'll have a list up shortly while the group 
decides where to host the project. If you want to be on the list, send 
me an email saying so and you'll get an introductory summary in 
response.

>
> - If I understand you correctly, you don't want to build Skate on top
>  of CLIM.  I understand that the higher levels of CLIM may be different
>  enough from Skate that it makes no sense to use them, but is there a
>  reason for not using at least the low-level CLIM stuff?  And if you're
>  not going to use CLIM for this, what _are_ you going to use?

I'm not making the decision; the Skate group will decide. But CLIM 
doesn't feel right to me for implementing Skate; some of what SK8 did 
(and Skate will do) is very low-level pixel-bashing; some of it is 
low-level policy construction, and CLIM has a pretty clear idea of its 
policy already; and free implementations of CLIM are not there on all 
the target platforms.

If you want to make a case otherwise, I invite you to join the Skate 
list and present your arguments; right now is just the right time for 
such discussions.

--me

In-Reply-To: <opshivruz3k6vmsw@...>
   References: <cnaqff$f1l$1@...> <8765466ghw.fsf@...> <41990B8E.3050408@...> <opshivruz3k6vmsw@...>
   Cc: mikel evins <mikel@...>, sbcl-help@...,
	   cmucl-help@...
   From: mikel evins <mikel@...>
   Date: Mon, 15 Nov 2004 13:42:39 -0800
   Sender: cmucl-help-bounce@...



   On Nov 15, 2004, at 12:12 PM, Arthur Lemmens wrote:

   > Mikel <mikel@...> wrote:
   >
   >> Again, the big missing features seem to be Unicode support and
   >> Win32 support.
   >
   > If "Win32 support" means "runs on Win32 systems", then I think CLisp
   > has both Unicode and Win32 support.  If "Win32 support" means "a usable
   > interface to a reasonable subset of the Win32 API", my impression is
   > that no open source Lisp has this.  (But I've never used CLisp or SBCL
   > or CMUCL, so please ignore me if I'm talking nonsense.)

   What it means is, 'runs on Win32 and provides a means of calling native 
   Win32 APIs.'

   There is some concern about the speed of CLISP. We did some kind of 
   hairy things on the original SK8 to make sure that certain things were 
   fast.

   Turns out that CMUCL does run on Win32, it's just that the version that 
   does is not yet released. We are in active conversations with the 
   person doing the port, and it looks like the features we need are 
   coming soon.

   --me


   In-Reply-To: <opshivruz3k6vmsw@...>
   References: <cnaqff$f1l$1@...> <8765466ghw.fsf@...> <41990B8E.3050408@...> <opshivruz3k6vmsw@...>
   Cc: mikel evins <mikel@...>, sbcl-help@...,
	   cmucl-help@...
   From: mikel evins <mikel@...>
   Date: Mon, 15 Nov 2004 13:42:39 -0800
   Sender: cmucl-help-bounce@...



   On Nov 15, 2004, at 12:12 PM, Arthur Lemmens wrote:

   > Mikel <mikel@...> wrote:
   >
   >> Again, the big missing features seem to be Unicode support and
   >> Win32 support.
   >
   > If "Win32 support" means "runs on Win32 systems", then I think CLisp
   > has both Unicode and Win32 support.  If "Win32 support" means "a usable
   > interface to a reasonable subset of the Win32 API", my impression is
   > that no open source Lisp has this.  (But I've never used CLisp or SBCL
   > or CMUCL, so please ignore me if I'm talking nonsense.)

   What it means is, 'runs on Win32 and provides a means 
   of calling native Win32 APIs.'

How about: Corman, GCL, and ECL?

I would also hope that it runs on the commercial Lisp systems.
(I use Xanalys Lispworks.)

   There is some concern about the speed of CLISP. We did some kind of hairy
   things on the original SK8 to make sure that certain things were fast.

I've heard that optimizing for CLISP is different than the kind
of guesses that people usually make for compiled Lisp code.

On Nov 15, 2004, at 2:00 PM, Christopher C. Stacy wrote:

>    In-Reply-To: <opshivruz3k6vmsw@...>
>    References: <cnaqff$f1l$1@...> 
> <8765466ghw.fsf@...> <41990B8E.3050408@...> 
> <opshivruz3k6vmsw@...>
>    Cc: mikel evins <mikel@...>, sbcl-help@...,
> 	   cmucl-help@...
>    From: mikel evins <mikel@...>
>    Date: Mon, 15 Nov 2004 13:42:39 -0800
>    Sender: cmucl-help-bounce@...
>
>
>
>    On Nov 15, 2004, at 12:12 PM, Arthur Lemmens wrote:
>
>> Mikel <mikel@...> wrote:
>>
>>> Again, the big missing features seem to be Unicode support and
>>> Win32 support.
>>
>> If "Win32 support" means "runs on Win32 systems", then I think CLisp
>> has both Unicode and Win32 support.  If "Win32 support" means "a 
>> usable
>> interface to a reasonable subset of the Win32 API", my impression is
>> that no open source Lisp has this.  (But I've never used CLisp or SBCL
>> or CMUCL, so please ignore me if I'm talking nonsense.)
>
>    What it means is, 'runs on Win32 and provides a means of calling 
> native
>    Win32 APIs.'
>
>    There is some concern about the speed of CLISP. We did some kind of
>    hairy things on the original SK8 to make sure that certain things 
> were
>    fast.
>
>    Turns out that CMUCL does run on Win32, it's just that the version 
> that
>    does is not yet released. We are in active conversations with the
>    person doing the port, and it looks like the features we need are
>    coming soon.
>
>    --me
>
>
>    In-Reply-To: <opshivruz3k6vmsw@...>
>    References: <cnaqff$f1l$1@...> 
> <8765466ghw.fsf@...> <41990B8E.3050408@...> 
> <opshivruz3k6vmsw@...>
>    Cc: mikel evins <mikel@...>, sbcl-help@...,
> 	   cmucl-help@...
>    From: mikel evins <mikel@...>
>    Date: Mon, 15 Nov 2004 13:42:39 -0800
>    Sender: cmucl-help-bounce@...
>
>
>
>    On Nov 15, 2004, at 12:12 PM, Arthur Lemmens wrote:
>
>> Mikel <mikel@...> wrote:
>>
>>> Again, the big missing features seem to be Unicode support and
>>> Win32 support.
>>
>> If "Win32 support" means "runs on Win32 systems", then I think CLisp
>> has both Unicode and Win32 support.  If "Win32 support" means "a 
>> usable
>> interface to a reasonable subset of the Win32 API", my impression is
>> that no open source Lisp has this.  (But I've never used CLisp or SBCL
>> or CMUCL, so please ignore me if I'm talking nonsense.)
>
>    What it means is, 'runs on Win32 and provides a means
>    of calling native Win32 APIs.'
>
> How about: Corman, GCL, and ECL?

They're under discussion.

>
> I would also hope that it runs on the commercial Lisp systems.
> (I use Xanalys Lispworks.)

So would I. As I said, though, more Lisps == more missing features to 
implement. It would be good to get some basic level of function working 
on one Lisp, to begin with.

>
>    There is some concern about the speed of CLISP. We did some kind of 
> hairy
>    things on the original SK8 to make sure that certain things were 
> fast.
>
> I've heard that optimizing for CLISP is different than the kind
> of guesses that people usually make for compiled Lisp code.
>
>

It may be so. We'll talk it over.

--me

> * mikel evins <zvxry@...> [2004-11-15 13:46:48 -0800]:
>
> Yes; I am soliciting discussion on the group mailing list about
> whether CLISP can perform well enough.

In that case, it might make sense to ask a CLISP list rather than
CMUCL/SBCL.

In my experience, CLISP is perfectly adequate performance-wise.
(OTOH, I use it for processing huge datasets on AMD64 which cannot be
handled by 32-bit CMUCL due to addressable memory size.)

> In the old days of SK8, certain functions were sometimes written in
> LAP to make them fast enough, so there is some skepticism about a
> bytecode implementation.

CLISP modules offer an easy way to write speed-critical functionality in
C/C++/Fortran/assembly &c


-- 
Sam Steingold (http://www.podval.org/~sds) running w2k
<http://www.camera.org> <http://www.iris.org.il> <http://www.memri.org/>
<http://www.mideasttruth.com/> <http://www.honestreporting.com>
Please wait, MS Windows are preparing the blue screen of death.

mikel wrote:
> 
> Again, the big missing features seem to be Unicode support and Win32 
> support. SBCL has the first, and CMUCL seems about to have the second. 
> Any other thoughts anyone cares to share would be welcome.

CMUCL did have support for unicode about 2 years ago or so. 
Unfortunately, it was never part of the main code so it's rotted quite a 
bit.  I tried briefly to bring it up-to-date, but being illiterate and 
not having any need for unicode myself, I gave up.  But perhaps this is 
motivation to bring it up-to-date.

Ray

>>>>> "me" == mikel evins <mikel@...> writes:
[...]
    me> If you want to be on the list, send me an email saying so and
    me> you'll get an introductory summary in response. [...]

Is it possible to allow gmane to carry the list whne one exists?  It
is possible to allow posting only for specific people so that need to be 
a concern. (lisp-hug works this way)

BTW, IMHO, what this thread has become makes it better suited 
comp.lang.lisp.  Could we move it there?  (since all participants seem 
to also read cll).

cheers,

BM