Don't be nice if this big company sponsor Squeak ?

http://research.sun.com/projects/squawk/

Any comments ?


__________________________________________________
Correo Yahoo!
Espacio para todos tus mensajes, antivirus y antispam ¡gratis!
¡Abrí tu cuenta ya! - http://correo.yahoo.com.ar

Edgar J. De Cleene wrote:
> http://research.sun.com/projects/squawk/
>
> Any comments ?

Wonder where this comes from ;-)
See the people page under the letter "I" :-)

Michael

On Nov 14, 2006, at 10:36 , Michael Rueger wrote:

> Edgar J. De Cleene wrote:
>> http://research.sun.com/projects/squawk/
>> Any comments ?
>
> Wonder where this comes from ;-)
> See the people page under the letter "I" :-)

Actually I think this started way before Dan went to Sun.

Sounds impressive though:

"The static footprint of the core system (interpreter, a RAM garbage  
collector, an non-volatile memory garbage collector), compiled from  
C, is 25 kB on x86.  The minimum runtime footprint in RAM is 520  
bytes for the Java heap and 532 bytes for native stack and data (on  
x86)."

- Bert -

Hi all (and Dan)!

Bert Freudenberg <[hidden email]> wrote:
I noted the name Eric Arsenau, didn't he write Pocket Smalltalk for the
Palm? And David Simmons also is mentioned somewhere in there.

regards, Göran

On Tue, 14 Nov 2006 11:47:04 +0100, Bert Freudenberg wrote:
> On Nov 14, 2006, at 10:36 , Michael Rueger wrote:
>> Edgar J. De Cleene wrote:
>>> http://research.sun.com/projects/squawk/
>>> Any comments ?

Sure: very nice find :)

Compare to www.superwaba.org (Open Source):

- SuperWaba.PRC 66 KB VM+GC (Palm platform, minus 7 KB launcher & icon)
- SWNatives.PRC 59 KB primitives (cross+plat.glue, minus 7 KB launcher &  
icon)
- SuperWaba!.PDB 204 KB class files (cross)

plus a litte St app  
(http://www.smalltalk.org/versions/LittleSmalltalk.html )

- thisSmalltalk!.PDB 56 KB (ported by me :)

The .PDB files have a dictionary section (all classes' full names) which  
is redundant.

/Klaus

> - Bert -

Oops, forgot to mention:

- image_version_0_9_bits!.PDB 47 KB (.image file)
- image_version_0_9_srce!.PDB 59 KB (.sources file & .changes file)

The .image file is just specs for (meta)classes, mDicts and methods  
(literals and bytecode). This can be substituted by full  
serialization/deserialization (not implemented due to my lazyness). So  
thisSmalltalk VM creates "native" 1st class LittleSmalltalk objects out of  
the specs, on launch (launch on my Tungsten|C (non-arm VM) is as fast as  
Squeak3.9 launch on my 1GHz notebook ;-)

The app offers a workspace with doIt+printIt and a classic browser.

/Klaus

On Tue, 14 Nov 2006 12:51:09 +0100, I wrote:

Sounds like they are writing a metacompiled language with low-level primitives capable of accessing the hardware directly. IMHO Sun should have done this fifteen years ago when they had a leg-up on the industry with Forth. Now, they are going to try again using a Java syntax. This sounds silly to me, as Java is just a "simplified" version of the very languages the article says they are trying to replace!

The only thing this will have in common with Squeak is that they both have a VM. The writers could just as well have compared Squawk to .Net, or, uh, Java!

2006/11/14, Another Dave <[hidden email]>:
Isn't the OVM (realtime Java VM) implemented it Java?

Philippe

Sorry, I'm not always up to date on my Squeak email...

>On Nov 14, 2006, at 10:36 , Michael Rueger wrote:
>
>>Edgar J. De Cleene wrote:
>>>http://research.sun.com/projects/squawk/
>>>Any comments ?
>>
>>Wonder where this comes from ;-)
>>See the people page under the letter "I" :-)

and Bert replied...
>Actually I think this started way before Dan went to Sun.

Yes, it did, but the name is no accident.  Squeak's compactness and simplicity and the write-in-itself-then-translate-to-C-for-deployment approach inspired Nik Shaylor's design that was carried out by the early team.

>Sounds impressive though:
>
>"The static footprint of the core system (interpreter, a RAM garbage collector, an non-volatile memory garbage collector), compiled from C, is 25 kB on x86.  The minimum runtime footprint in RAM is 520 bytes for the Java heap and 532 bytes for native stack and data (on x86)."

These are the stats for the early version;  the current implementation that supports the SPOTs is somewhat larger, but it has lots of other (cool) features, most especially all the mechanism for wireless mesh networking.  There's lots more about that in various papers about the SPOTs.

An especially interesting aspect of the current challenge for Squawk is performance.  In the world of SPOTs, the most important figure of merit is power, so instead of the speed issue (nanoseconds per bytecode) you have a power issue (picojoules per bytecode).  And space is also an issue, of course, so you get these interesting strategies such as compiling ahead of time, then compressing for space, and storing in flash.  Here the analogy to JIT is decompressing out of flash into the RAM working set.  Which is better can be a matter of which uses less energy!

I did oversee the Squawk project for about six months after arriving at Sun and before starting my current project, but all the work was done by the other team members listed on the Sun site.

        - Dan

On Tue, 14 Nov 2006 12:54:55 +0200 , [hidden email] wrote:

> I noted the name Eric Arsenau, didn't he write Pocket Smalltalk for the
> Palm? And David Simmons also is mentioned somewhere in there.

Eric didn't write Pocket Smalltalk, but he did take over the maintenance of
it.

Eric is a long time Smalltalker (he was using ST-80 on a Tektronix 4044
back in 1987 when I first learned Smalltalk), although he works for Sun now
doing Java stuff, he is definitely motivated by his Smalltalk background in
his Squawk work.

Later,
Jon

--------------------------------------------------------------
   Jon Hylands      [hidden email]      http://www.huv.com/jon

  Project: Micro Seeker (Micro Autonomous Underwater Vehicle)
           http://www.huv.com

Hi everyone,

I chatted with Eric this morning (we are old friends), and he sent this to
me, to post for him (after reading the thread on Yahoo Groups):

(and, btw, Eric reminded me it was a Tek 4404/4406, not a 4044)

==================================================

On Wed, 15 Nov 2006 08:13:36 -0800, Eric Arseneau <[hidden email]>
wrote:

You guys have no idea what it was like seeing Dan Ingalls walk out of a
Sun building.  I thought I had stepped into the twilight zone :)  Then
to find out he worked in Sun Labs on some really cool stuff, I could not
resist finding a way to work for him.

When I did, it turned out to be for this Squawk project.  This project
also turns out to be quite interesting in itself.  As Dan points out it
is targeted towards resource constrained devices and is/was intended to
be modeled after Squeak.  In fact, check out the attached logo I came up
with for fun.  But Dan wont let me use until we come to the full
realization of generating the entire VM from Java (we are missing the
interpreter at the moment due to some shortcut taken a year or two ago).

And as Jon pointed out, I may not have written Pocket Smalltalk, I did
do a significant amount of work with Andrew Brault (the original author)
on the final version and some the Squeak version which was never
completely released. I did however write Pocket Java, which was Java
running on top of Pocket Smalltalk.  Andrew thought I was a complete nut
for doing this, when I could work on such a beatiful language as
Smalltalk :)  Here are the slides from the Smalltalk Solutions 99, where
Andrew and I presented this stuff
http://www.pocketsmalltalk.com/Solutions99/index.htm.

One more comment, to answer Another Dave
"Now, they are going to try again using a Java syntax. This sounds
silly to me, as Java is just a "simplified" version of the very languages
the article says they are trying to replace!

The only thing this will have in common with Squeak is that they both have
a
VM. The writers could just as well have compared Squawk to .Net, or, uh,
Java!"

You are sort of correct on the syntax part, however you miss the point
that Java is much more than its syntax, it is its tools, its libraries
and accessibility that are more important.  Just like Smalltalk,
although the Smalltalk syntax does speak for itself quite well :)  Being
able to leverage those things does make implementing the VM in Java
useful to a number of people who may or may not be comfortable with
building a VM.  Never mind all the cool stuff you can do using a real
environment over C/C++.  And the commonality is the ideology behind
Squawk is to write the VM in its own language, as well as write most of
it in itself, reducing the number of necessary primitives.  Up until a
few months back, the GC was written AND executed in Java by the
interpreter.  This is concept that Java people had not really contended
with before and it really throws them for a loop :)

Eric Arseneau

==================================================

--------------------------------------------------------------
   Jon Hylands      [hidden email]      http://www.huv.com/jon

  Project: Micro Seeker (Micro Autonomous Underwater Vehicle)
           http://www.huv.com