An OLPC Development Model

[Samuel Klein]

On Thu, May 8, 2008 at 11:55 AM, C. Scott Ananian

ps. SJ, there are no 'core activities' that we ship.  There is only
> one security-privileged activity (Journal), which we currently ship in the
> core build because (a) Sugar breaks otherwise, and (b) Rainbow's
> activity-signing stuff is incomplete.  I hope we can fix both of these
> in time, and stabilize the APIs enough that we can eventually unbundle even
> Journal.


Just fine.  We have more work to do to comply with the GPL in this case...
A build that loads Sugar without successfully loading a bundle stream should
give visible indications that it is not yet complete.


> Your notion of 'core activity' is probably worthwhile
> for support and documentation reasons
>

There are many related notions of coreness.  While I wasn't trying to define
a new one, in the context of feeds and activity status tags, I will offer
one:  the intersection of "globally supported" "important to education"
"localized into the ambient language" and "moderately demanding" (in size
and resource consumption).  This could be one of the default available
streams, suitable for priming all machines in a region, and further tweaked
by teachers and students on receipt.

Preparing / testing / localizing / updating the latest "supported" version
of an activity has its own timeframe; smooth deployment depends on all
parties involved planning sufficiently far out.   So while your defintion of
'build process' is one which excludes  activity development, there is a
parallel build process for named activity-sets which we should be talking
about with clarity.

SJ
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.laptop.org/pipermail/devel/attachments/20080508/93e44bce/attachment.html>