d***@avidriswold.mailshell.com
2006-09-12 01:07:48 UTC
For those of you who have heard of Strongtalk (or if you haven't, but
you care about performance in Smalltalk, or type systems), there is
great news today:
Sun Microsystems has at long last released the Strongtalk
virtual-machine as Open Source! This means the entire Strongtalk
project is now fully Open Source, under a basically unrestricted
Berkeley-style license, at http://www.strongtalk.org/.
As many of you may remember, Strongtalk was (and still is) the fastest
implementation of Smalltalk ever. It also included the only fully
developed type-system for Smalltalk (use of which is optional, and
doesn't affect performance). The VM is based on Urs Hölzle's highly
advanced type-feedback compiler technology, from the Self project at
Sun Labs. It was developed in the mid-'90s by a startup, Animorphic
Systems, a startup company, but was shelved when Sun bought Animorphic.
At Sun the Animorphic team based the design of the current (HotSpot)
Java VM on the Strongtalk VM. But Strongtalk sat on a shelf, not
quite finished.
In 2002, Sun released Strongtalk with a non-commercial binary of the
unfinished VM, with open source just for the associated
Smalltalk/Strongtalk libraries. But without the VM source code to
allow finishing and maintaining the VM, Strongtalk couldn't go
anywhere, other than as a historical proof that you really can make
Smalltalk run very fast, and that you can put a static type-system in
Smalltalk without ruining the feel of Smalltalk.
Now, with the release of the VM source code, there's nothing stopping
anyone from doing anything with the Strongtalk system, and using it for
basically any purpose. This would also open the possibility that other
Smalltalk implementations could borrow from the technology. To whet
your appetite, imagine something like Squeak running on a VM that is
many times faster. We don't want to compete with other Smalltalk VMs-
we just want Smalltalk to run a lot faster, however that has to happen.
To keep our feet on the ground, this is a very sophisticated, complex
system that is not fully finished, so there is a huge amount of work
required, by people with deep knowledge of VMs, to finish it or adapt
the technology to VM designs. But before it was impossible. Now,
what happens is entirely up to the Smalltalk community!
Many thanks to Gilad Bracha at Sun for persevering and making this
happen, to Sun for releasing it, and to Robert Griesemer for helping
figure out how to get the VM building again after 10 years on a shelf.
-The Strongtalk Team
you care about performance in Smalltalk, or type systems), there is
great news today:
Sun Microsystems has at long last released the Strongtalk
virtual-machine as Open Source! This means the entire Strongtalk
project is now fully Open Source, under a basically unrestricted
Berkeley-style license, at http://www.strongtalk.org/.
As many of you may remember, Strongtalk was (and still is) the fastest
implementation of Smalltalk ever. It also included the only fully
developed type-system for Smalltalk (use of which is optional, and
doesn't affect performance). The VM is based on Urs Hölzle's highly
advanced type-feedback compiler technology, from the Self project at
Sun Labs. It was developed in the mid-'90s by a startup, Animorphic
Systems, a startup company, but was shelved when Sun bought Animorphic.
At Sun the Animorphic team based the design of the current (HotSpot)
Java VM on the Strongtalk VM. But Strongtalk sat on a shelf, not
quite finished.
In 2002, Sun released Strongtalk with a non-commercial binary of the
unfinished VM, with open source just for the associated
Smalltalk/Strongtalk libraries. But without the VM source code to
allow finishing and maintaining the VM, Strongtalk couldn't go
anywhere, other than as a historical proof that you really can make
Smalltalk run very fast, and that you can put a static type-system in
Smalltalk without ruining the feel of Smalltalk.
Now, with the release of the VM source code, there's nothing stopping
anyone from doing anything with the Strongtalk system, and using it for
basically any purpose. This would also open the possibility that other
Smalltalk implementations could borrow from the technology. To whet
your appetite, imagine something like Squeak running on a VM that is
many times faster. We don't want to compete with other Smalltalk VMs-
we just want Smalltalk to run a lot faster, however that has to happen.
To keep our feet on the ground, this is a very sophisticated, complex
system that is not fully finished, so there is a huge amount of work
required, by people with deep knowledge of VMs, to finish it or adapt
the technology to VM designs. But before it was impossible. Now,
what happens is entirely up to the Smalltalk community!
Many thanks to Gilad Bracha at Sun for persevering and making this
happen, to Sun for releasing it, and to Robert Griesemer for helping
figure out how to get the VM building again after 10 years on a shelf.
-The Strongtalk Team