Android, OLPC, and native hosting

[NoiseEHC]

>> Actually, no. The .class -> .dex compiler consumes an enormous amount of
>> memory, so it is out of the question at least for now.
>
> How much is enormous ?  A laptop/tablet is likely to have more than
> a smartphone...
>

With hundreds of classes in a .jar to convert it uses some 256M, with 
thousands it uses more than 1GB... (Likely it will be optimized in the 
future though.) Another problem is that those development tools use such 
native APIs which are not supported by the Android NDK (Native DevKit). 
So either those tools (like the java runtime or the make program) should 
be ported to the NDK (but why waste so much effort on this?) or the 
development environment should be installed in the system image. The 
latter one just wastes flash and probably opens up some nice security holes.

>> However what I do not get is why would it be good for an education 
>> project if it would be
>> self hosting at all? As I see an education project's main goal is to
>> support creating quality educational resources (like curricula) cheaply,
>> is not it?
>
> You can't deny kids the ability to create their own activities 
> (applications, whatever)
> Pippy is an example of a simple way to introduce kids to activity 
> programming in
> python, allowing them to easily create and share activities.

You can still create applications with
http://code.google.com/p/android-scripting/

With the existing tools it is true that children cannot create the same 
quality applications what is possible with the Android SDK environment 
(even if we include a ported Etoys, TurtleArt and Karma), but they could 
create and share applications with the currently existing tools so what 
is the problem? The problem with self hosted activity development is the 
nonexistent development environment rather than the limited 
functionality. Even if the compilers will be ported to the Android NDK, 
Eclipse will never be ported so programming Android on the XO-3 (or 
XO-1.7) will be just as painful as programming Sugar with Pippy today. A 
much more simple solution would be just shipping a full fledged Linux PC 
to every school and let children log into it with VNC. So the ~3% of 
children who can become programmers would be able to develop the same 
applications (with Eclipse) what we can and the rest of the children 
would just use some simplified environment like scripting...