Spoon July 2003

spoon@lists.squeakfoundation.org

2 participants
4 discussions

Squat progress 19 July 2003: new version of complete installation package
by Craig Latta 20 Jul '03

20 Jul '03

(mailing list info at http://netjam.org/squat ) Hi-- Now that I've made a suitably small virtual machine, I've finished a new version of a complete installation package for Squat, along the lines of "relief" (http://netjam.org/relief ). It's a webpage that, when visited, downloads an installer program and runs it. The installer program downloads a snapshot, a virtual machine, and the Flow plugin, decompresses the virtual machine and the Flow plugin, and runs the virtual machine. The virtual machine decompresses the snapshot, and the snapshot updates itself. The current sizes are 93,352 bytes for the snapshot, 18,973 bytes for the Flow plugin, 107,642 bytes for the virtual machine, and 61,440 bytes for the installer program, for a total of 281,407 bytes. The decompressor is is both the virtual machine and installer; I could save about 40K by loading it dynamically in each case... I'll save that for some future version. Next I'm going to write my first Squat application, which is a snapshot that updates itself enough to display my resume, then displays it. :) thanks, -C -- Craig Latta http://netjam.org/resume craig(a)netjam.org [|] Proceed for Truth!

1 0

Squat progress 17 July 2003: snapshot work finished, VM work started
by Craig Latta 19 Jul '03

19 Jul '03

(mailing list info at http://netjam.org/squat ) Hi-- Today I unloaded all the compiled methods which hadn't been run since just before the previous snapshot, made a snapshot, and compressed it with bz2. I added a bz2 decompressor to the virtual machine, and successfully ran the snapshot. The snapshot size on disk is 93,352 bytes ("91 KB"). I consider the snapshot part of the project finished, although I'm sure a few more bytes could be eked out by doing symbol compression, etc. (I'm not going to bother, though, since they make using the snapshot significantly less pleasant). There is no accurate space analysis for this snapshot, since the act of making one would change the size of the snapshot significantly. :) A space analyzer that ran from a simulator would be useful. When the snapshot is decompressed into RAM, and a space analysis is printed, the space analysis at http://netjam.org/squat/space2.txt is accurate. In summary, the 91 KB file yields a 280 KB object memory, and the system is arbitrarily extensible from there. Now I'm working on making the smallest possible virtual machine that will run this snapshot. thanks, -C -- Craig Latta http://netjam.org/resume craig(a)netjam.org [|] Proceed for Truth!

1 0

Squat progress 17 July 2003: method activation marking
by Craig Latta 18 Jul '03

18 Jul '03

(mailing list info at http://netjam.org/squat ) Hi-- I've added support for marking each method as it is run, giving me a better (but not perfect) idea of which methods I can swap out. I've got a few interesting numbers. If I clear all method marks, then snapshot, resume, and ask for a tally of marked and unmarked methods, I get 458 marked and 1766 unmarked (a full space analysis is at http://netjam.org/squat/space2.txt ). If I clear all method marks and immediately ask for the tally I get 27 marked and 2197 unmarked. (The queries are all done via remote message-sending, to a snapshot which has no methods swapped out.) -C -- Craig Latta http://netjam.org/resume craig(a)netjam.org [|] Proceed for Truth!

1 0

Squat progress 12 July 2003
by Craig Latta 14 Jul '03

14 Jul '03

(Squat mailing list info at http://netjam.org/squat ) Hi-- Things are getting exciting now. :) I can now swap out compiled methods and method dictionaries on demand, then automatically and seamlessly swap them back in from an update server when messages requiring them are sent. This ability provides for a significant reduction in image size (about 30%). I'm in the process of swapping out most of the behavior in the minimal system. Space analyses will be at http://netjam.org/squat/space2.txt . I have a demo where I open a remote browser on a headless snapshot that has had all the compilation and decompilation support removed, then, as I browse around, some of it gets swapped back in. I plan to make a short movie of this. I also have a design for writing snapshots to a socket, allowing the removal of all filesystem support. A list of projected surviving classes is at http://netjam.org/squat/survivors . thanks, -C -- Craig Latta http://netjam.org/resume craig(a)netjam.org [|] Proceed for Truth!

2 2

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

Spoon July 2003