scheduling squat (was "draft rough plan for 3.6")
craig.latta at netjam.org
Mon Apr 14 20:55:33 UTC 2003
> Is there any information on what the Squat project entails? I've just
> been talking to someone who is interested in "small environments" and
> I would like to point to Squat and what's being done in there.
The best source of information currently is the Squat mailing list
archive, which has a couple of progress reports. Please see
http://netjam.org/squat/. I sent the most recent report to the Squat
list and to the Squeak-dev list, and I've repeated it below.
> What's the relation between 4.0 and Squat (if any)?
Only that I think Squat is something to schedule for a major release,
not a minor one.
> Last time we chatted (the other day) [Craig] was down to around 400kb
> and was going for 200kb.
Actually, I don't think I said 200k. :) I'm aiming for 100k.
> ...[Craig] thought it would be too soon in 3.6 - but potentially
> sometime after.
Again, it's not that I think it would be too soon, just that it would
be too fundamental a change for a minor release.
Squat progress, 26 March 2003
Craig Latta craig.latta at netjam.org
Wed, 26 Mar 2003 16:22:23 -0800
In my last update, I mentioned that I had installed a minimal version
of Flow into a small (600k) snapshot. I am now communicating with that
snapshot from a normal one, via sockets. I have also implemented a
minimal remote message-sending facility (AKA an "ORB"), so that I can
send messages from the normal system to the small one via proxies, with
complete transparency. (I didn't use one of the extant ORBs for various
technical reasons, see postscript below.)
I'm using the remote-sending facility to direct the further shrinking
of the small snapshot from the normal one. This way, it's very
straightforward to shut down the human interface, for example. I'll
probably do some final cleanup with the simulator (I've added support
for making snapshots, along with a few other utilities; among other
things, I've made a small snapshot with the GUI ripped out via
simulation). I also plan to use remote-sending to build up snapshots
from the minimal one.
None of the Squeak ORBs I looked at (Diego Gomez Deck's rST, Xu Wang's
S2S, John Fabry's RMP, Peter Smet's FlyByWire, Masashi Umezawa's
SOAPOpera, Lex Spoon's RemoteObject framework) were quite the right
thing for what I'm doing. Generally, they didn't provide complete
forwarding coverage. What I've done forwards absolutely everything, even
messages normally handled by the virtual machine, or resolved in the
method cache (the proxy class I've made won't even compile messages sent
to self in instance methods :). I don't use >>doesNotUnderstand: (in
fact, that would get forwarded :), and the proxy class can be inserted
anywhere in the class hierarchy. It's also decently fast, and simple.
craig at netjam.org
Smalltalkers do: [:it | All with: Class, (And love: it)]
More information about the Squeak-dev