On Tue, Aug 21, 2007 at 07:11:57PM -0700, Steve Lamb wrote: > Not that all of the libraries need to be moved into the root partition > for it to function as a stable init platform. Outside of sys (which is > compiled in), os, re, shutil, time and datetime I can't think of anything > that is needed for scripts running at init time. Certainly not many of the > extended Python libraries like a minimal web server, a whole suite for > handling email and MIME, TK/TCL bindings and so on. A very rough check and > I think 2.5Mb would be sufficient to serve as an init platform and natively > cover more than 80-90% of any operation an init script would need. So that > increase of 10-20% at 10Mb is more like 3-5%, if that. Neat. Anyone have a testbed to try it on? I mean, move the minimal python stuff to /bin and /lib and remove it from /usr, verify that it runs, then try replacing some initscripts? I'm all out of spare boxes right now.
Taken to the nth degree, it could totally change how the initscripts work. Instead of separate scripts in /etc/init.d run one at a time by run-parts, they could be individual modules (and hense optimized), run by a new /etc/init.d/rc python script. This would start off by importing each module then running its main function. This would totally do away with the process overhead of the sysVinit system while maintaining support for run levels. However, since unixes are in part differentiated in how they handle init scripts, replacing sysVinit with PythonInit would be a huge deal. At a minimum, it would require a new update-rc.d command so that packages would be setup correctly. Update-rc.d would also have to detect if a package included an sh startup script and setup the PythonInit to os.system the script instead of loading it as a module. Doug. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
