[Community-news] OLPC News 2007-09-22

[Walter Bender]

1. David Cavallo has formed a learning team that will work with
countries to develop their own learning teams. The goal is to help
each country develop a deeper understanding and richer practice in
using laptops for learning. We will hold workshops at OLPC each month,
work in countries, and collaborate internationally. The first
workshops in countries will be in Ethiopia, Ghana, Rwanda and Nigeria
during October.

2. Schedule: We still have not gotten to a full code freeze as there
are too many open critical bugs. If you have blocking or high-priority
bugs open for Trial-3, please help us by analyzing the work load,
making a suggestion for a work-around or how to document the issue,
and setting the bug to "untriaged" so that the triage team will review
it. Please check with Jim Gettys or Kim Quirk before checking in any
code changes.

3. Test: Alex Latham updated the test configuration notes in the wiki
to address network-access configuration scenarios: two laptops under a
tree; laptop connected to an access point; and laptop connected to a
school server mesh. There are also some additional configuration notes
and test plans for localization (See
http://wiki.laptop.org/go/Test_issues). Alex Khitrick and Yani Galanis
have both "joined" the test team this week to help with executing test
cases, debugging, and writing up bugs.

4. Hardware: The hardware team (which includes people from Quanta and
AMD) spent this week trying to track down some of the remaining
problems with suspend/resume. We have reached a point where many
laptops are running for hundreds of thousands of cycles without
crashing, but occasional crashes still occur (and some laptops are
more susceptible than others.) No other laptop comes close; but
neither do they have our ambition to suspending between each page read
(or maybe even suspending between keystrokes), with potentially
thousands of page views/day.

The low noise margin of the +3.3V line that powers most inter-chip
communication (and the WLAN) on the laptop was the big surprise. An
even bigger surprise was that the margin was critically dependent on
battery voltage! When the battery was low, our 3.3V supply was
dropping to below 3.15V, due to insufficiently turning on the
transistor used to switch +3.3V off during suspend.

John Watlington and Quanta have a proposed solution  which is being
tested. We hope that improving this will close out the "unable to
resume via power button or wireless" bug completely (See
http://dev.laptop.org/ticket/1835)—it was actually two distinct
problems, one of which was fixed last week—and we have reason to hope
that it might be at least partially responsible for the "USB wireless
suspend/resume failure at setup phase" bugs (See
http://dev.laptop.org/ticket/1752 and 2621).

Note: anyone working on a suspend/resume problem should keep their
machine plugged into line power until it has been properly modified.

Chris Ball ran overnight tests on many of the proposed hardware
changes. The wireless resume bugs are now occurring rarely enough to
have taken a backseat to other kernel bugs related to resume (See
http://dev.laptop.org/ticket/3477 and 3479).

5. Firmware: Mitch Bradley released Q2C27, the firmware released
planned for Trial-3. It includes secure boot capability (but can also
boot in a non-secure fashion). It has been tested on B1, B2-1, pre-B3,
B3, and B4 systems. Do not use it on pre-B1 boards—it will brick them
because of changes to the embedded controller (EC) microcode. Also
note that it does not work on A-Test boards as support for A-Test has
been eliminated from the EC microcode.

6. EC code: Richard Smith re-factored the way the EC does its startup
procedure so that an automatic restart now happens after a reflash.
This involved re-working the way the IO maps (B2,B3,B4,C1) were
detected. Alas, the rework seems to have broken the power button;
when Richard gets that issue sorted out, we should be able to to fully
automatic firmware updates.

7. X Window System: Bernie Innocenti has been working mostly on visual
bugs this week. The "iGoogle" bug was hard to find and easy to fix.
The "green icons" bug was easy to find and hard to fix. Bernie has
also pushed the Ethiopian and Urdu fonts in the builds.

8. Security and updates: Scott Ananian and Michael Stone worked
closely together to implement the "live updates" feature (See
http://dev.laptop.org/ticket/2517). This feature, available since
Build 595, allows the laptop user to run the command "olpc-update NNN"
to update to Build NNN. Scott worked with Michael Stone to get an
initial in-system network upgrade in place. After completing live
updates, they discovered certain inefficiencies in our implementation
of manifest verification. Michael profiled the software and, with
Scott and Ivan Krstić's help, developed an enhancement that halves the
time required to verify a manifest. Scott improved the speed of the
in-place upgrade by 10× by rewriting contents verification.

Scott worked with Mitch Bradley on shaking out the secure-boot process.
Among other things, they tweaked the developer and activation lease
formats based on suggestions from Michael Stone and SJ Klein. Scott
also worked on creating a pure-Linux reimplementation of the
auto-reinstallation process, eliminating the two reboots into OFW and
making it more compatible with the secure-boot process.

9. Kernel: Andres Salomon spent the week testing open bugs; he helped
folks who were working on the i2c timeout issues, prepared patches for
upstream, reviewed lots of patches, and did a cleanup of the olpc-2.6
git repository. The git cleanup means a clone takes only 15 minutes,
down from 90 minutes.

10. Licensing content: Jon Phillips has been leading an effort to
integrate Creative Commons licenses into the Journal. Screen shots of
his mock ups are in the trac system (See
http://dev.laptop.org/ticket/3529).

11. Sound samples: Dr. Richard Boulanger continues to gather together
collections of sound samples recorded specifically for OLPC by the
team at Open Path Music. The most recent additions are are vocal
sounds: animal noises, grunts, groans, growls, etc., plus some
singing. These samples are being incorporated into activities such as
TamTam.

12. Library: Mako Hill helped rewrite the Javascript bits of the
library to use Python templates using the Jinja templating system (See
http://jinja.pocoo.org/). This is a more robust implementation if
Javascript is turned off. The templates will be reused once we turn on
a web server in the builds and change from a model of "update
bundle-list on download" to a dynamic view that checks the filesystem.

The default browser page will point to the local content repository
and have links to browse the web. Currently, there is are links to:
* google.com, which will auto-redirect to the proper language for the country;
* the school server;
* wiki.laptop.org/go/Home (initially English by default) from which
the activity and content bundle download pages will be more
prominently linked;
* wiki.clusty.com, to search Wikipedia (initially English by default)
Additional suggestions are welcome.

13. Biology activities: David Stang of ZipcodeZoo is working with
Lauren Klein to provide a kid-friendly interface to his database of
bug information for bug blitzes. Charles Smith at the EOWilson
foundation is working on basic ideas and background information for
classes running blitzes. Misha Herscu, a teenager who has already
published his first book on herpetology in Massachusetts, is helping
organize the work with zipcodezoo. Misha also helped connect the Thai
group that was doing their own bug-finding activity with an unusual
bug zoo in northern Thailand, who may be able to help them extend the
work they have done. and perhaps take some of the first online photos
of certain species (See http://www.malaeng.com/).

14. Schuyler Erle, working on mapping and radio projects for UNICEF, has
gotten a radio program to work on the XO, which can make the XO an
audio receiver and broadcaster.

-walter

-- 
Walter Bender
One Laptop per Child
http://laptop.org