Bob Courchaine wrote on Thu, 26 May 2005 14:12:12 -0500
Tim, in your opinion, would this be worth trying one more time?
This reminded me that I should have done a follow up to my "Squeak for MIT Media Lab laptop project" proposal a while ago.
http://lists.squeakfoundation.org/pipermail/squeak-dev/2005-April/090833 .html
When the end of April came and went with nobody joining the team I decided to consider it a failure and to retract my offer to lead such a group in whatever direction the Squeak community thought best. I did enjoy the related discussions on this list, however, and learned much from them. So much so that I am still doing this, but on my own and in the direction *I* think is best.
In this case I took Lex's "old code is good code" advice very seriously and came up with a way to start out with a modified Squeak and then build my Neo Smalltalk *under* it (no license worries ;-) piece by piece. I had always planned to include Squeak compatiblity, but the previous idea was to leave that for later and add it slowly.
Instead of running Squeak bytecodes directly the hardware will run Plurion bytecodes (an early draft can be found in http://www.merlintec.com/download/plurion.pdf) with a small software layer on top of that (which is later to grow into the full Neo Smalltalk) to support the Squeak bytecodes. The next layer up will be a small Squeak image including all of the virtual machine but the bytecode interpreter. This will probably be based on Spoon. The top layer will be a regular Squeak image.
Meanwhile, I have been trying to get something close to my original option 2 (Squeak on hidden Linux layer) running on my next web server machine. Fflinux seems too specific to the ITX motherboard, so that wasn't a good option. Other very small Linuxes seem to be based on 2.0.x kernels and that probably isn't a good idea for what I want to do. I tried the "Linux From Scratch" route but can't even compile glibc properly. Suddenly rewritting a whole TCP/IP stack in Smalltalk doesn't sound so bad ;-)
-- Jecel