[sugar] Squeak / Etoys RPMs

[Ian Bicking]

Ivan Krsti? wrote:
> Ian Bicking wrote:
>> E.g., the Cheetah template language has 420Kb of tests.  
> 
> We'd presumably drop those.
> 
>> My app has 300Kb of Javascript. 
> 
> We do have compression on JFFS2, so something like Javascript will
> certainly wind up quite a bit smaller on the actual flash.
> 
>> Paste includes 600Kb that is mostly backported
>> stdlib libraries in case you are running on Python 2.3.  I'm not
>> entirely sure how to deal with this issue.
> 
> I think it's reasonable to say that backwards-compatibility code should
> be absent from OLPC activities whenever such code can be relatively
> cleanly removed (i.e. isn't deeply embedded all over the codebase). No
> idea whether that's the case with Paste.

There's lots of specific ways to improve this one app.  The larger issue 
I'm wondering about is how to build smaller packages without forking 
everything.

Part of the reason the sizes are so large is because I -- and many other 
developers -- have been pretty free with disk space.  And why not, it's 
dirt cheap.  OLPC changes that.  So I wonder if we should think about 
ways to support lighter builds of packages, that may be acceptable 
upstream, or may not, but are maintained separately.  There's really 
just two settings that control that in a setup.py file, and we can 
override them specifically for the bdist_olpc_bundle command.

Also, should we consider leaving out .py files?  It pretty much doubles 
the size of the code.  While I like the idea of including proper source, 
for most students it won't be interesting at all, and perhaps it would 
be better to leave them out but make it easy to get the "full" activity 
(which in addition to .py files, might also have tests, which are also 
not interesting to people just using the activity).


-- 
Ian Bicking | ianb at colorstudy.com | http://blog.ianbicking.org