OLPC News 2008-02-02

[Walter Bender]

1. Active antennae: Another 90 prototype active antennae should be
available in a couple of weeks, followed shortly by a large shipment
of pre-build antennae scheduled to arrive in three or four weeks. The
initial run will be used mostly for field testing, with the majority
of the units going to Uruguay. They will be labeled as "engineering
samples—not for sale." We now have an update procedure for the
prototype antennae that allows them to stay connected to a server.
(These had been built with firmware that placed them in stand-alone
mesh-repeater mode too quickly, thus requiring them to be connected
only after a server is up and running.) See
http://wiki.laptop.org/go/Active_Antenna_Reprogramming.

2. Firmware: Mitch Bradley fixed a problem with OFW reading JFFS2
images (Ticket #6291) encountered when using the multicast update
method. (This was one of the bugs uncovered by David Woodhouse in
Mongolia last week.)

3. School server: Power continues to concern us. John Watlington
realized that the off-the-shelf server prototype he was looking at for
rural environments actually came with a 19VDC power supply, not a
12VDC one. While 12V supplies are available, they don't work well with
unregulated 12V input. With such a 12V supply, the server prototype
required around 16W while idling, and up to 26W when running three
meshes and doing heavy disk accesses. The current power consumption
requires four hours of pumping on a Weza to keep the server operating
for an eight hour day! We will also have to greatly improve the power
consumption when the machine is idle to have any hope of the servers
being left running when the schools aren't in session.

4. Embedded controller: Q2D10 had some battery charging regressions,
so Richard Smith backed out the change that speed up the
battery-processing state machine; that fixed the regressions. The EC
command saga continues: a  machine was brought in that had total EC
command failure, yet after Richard started examining it, it magically
cleared up. After a long spell of trying to reproduce the problem,
Richard made a significant discovery: it appears that if the
input-buffer-full (IBF) flag is set and the power to the processor is
cut, then the EC can go into a state where it thinks that a constant
stream of data is being received. This results in the IBF flag getting
reset just a soon as you clear it. Richard is still
researching/understanding the issue, but this may explain why the
previous interrupt-driven protocol was having so much trouble.

5. Automated charging testbed: Richard has set up an automated
charging testbed: four XO laptops are now in a suspend/resume testbed;
these laptops are connected to a switch such that every three hours, a
supervisor machine turns off the external power to each of them. Each
laptop is running a small script that watches for when the battery
capacity gets low. When low battery is detected the XO laptop turns
its power back on.

6. Power profiling: Now that we have automatic power management in the
Update.1 builds we no longer have a simple power profile for measuring
battery life. To get an accurate indication of what the "real world"
battery life will be when power management is doing automatic
suspend/resume we need to know what the power profile looks like while
using the machine. We are gathering data from different use cases by
running the olpc-logbat script while using the XO laptop: olpc-logbat
samples the battery discharge information every 10 seconds. We can use
much more data—please run the script yourself and send us the CSV
files that it generates.

7. Testing: Much thanks to Chih-Yu Chao, whose last full time day
helping with QA and testing was Friday. This week she was focused on
providing test cases, structure and encouragement to the community in
our push for Update.1 testing. To help out, please review and execute
test cases listed in the wiki (http://wiki.laptop.org/go/Update.1), or
choose some test plans (http://wiki.laptop.org/go/Category:Test_plans)
and then post the results
(http://wiki.laptop.org/go/Update.1#Test_Results). We can really use
lots of help!

Yani Galanis has been testing avahi, telepathy, and general mesh
capabilities with the latest Update1. He has helped open up some
discussions of what we have today, what we would like in the future,
and how we might get there. There is still some design work, coding,
testing, and discussion needed in this area as some of our real
deployments are pushing at our limitations.

8. Support: This week Nicholas Negroponte sent out a letter to all
donors who have not yet received their laptops apologizing for the
problems and explaining some of the on-going issues. The remaining
laptops should be shipped by the end of  March. Many people can now
track their order directly at the laptopgiving.org webpage, which has
started to reduce the number of emails to the support team.

There was a good discussion on Friday with Mel Chua, Nicki Lee, SJ
Klein, Adam Holt, Walter Bender, Kim Quirk on the topic of grass-roots
repair centers—more on that theme next weekend.

Adam organized another Sunday meeting among ~20 support volunteers,
with guest speaker Manusheel Gupta talking about entrepreneurship
among children with XO laptops. The ~60 support volunteers continued
to fend off shipping/billing questions this week by the 100s. The
number of questions almost doubled in January; the percentage of
questions pertaining to donor services increased four fold: almost a
700% increase from December!!

But there is lots of good news: even with the continuing onslaught
regarding donor services, we've lowered our unresolved tickets queue
from 500+ to about 350—and we have received many profuse thank-you
letters from donors who had been fed up to the gills at being
abandoned until now. Even with the increase in volume of laptops
deployed, there was no corresponding increase in questions about
connectivity, Flash support, or help getting started. It is safe to
say that once people get their XO laptops, they are managing quite
well.

Our thanks to dwa (David Aquilina) and alc (Alan Claver) and countless
other volunteers working so hard to be supportive towards all.

9. Roadshow: Dave Woodhouse, Bernie Innocenti, and Jim Gettys attended
Linux Conf Au (LCA), which is considered by many to be the best Linux
conference in the world at this time. The LCA organizers and OLPC
combined to distribute a 100 machines to developers at the conference.
(The lack of G1G1 in Australia is, of course, frustrating to people
here.)

10. Presence Service: Guillaume Desmottes ran more tests on Salut
using "hyperactiviy." He fix various memory leaks and other issues and
helped Marco Gritti to use hyperactivity to debug a Sugar UI bug. He
also reviewed Sjoerd Simons's Gibber DNS resolver branch; released
telepathy-salut 0.2.2, which fixes some OLPC related issues (Ticket
#6271); and discovered and tracked more Salut crashers (Tickets #6303,
#6309, #6310). Morgan Collett used his new Fedora superpowers to build
RPMs for Presence Service and Salut via koji. He has been working on
Presence documentation and tested builds and
presence service/telepathy related fixes. Sjoerd fixed several bugs in
telepathy-salut that were discovered thanks to the hyperactivity
stress testing tool.

11. Sugar: Simon Schampijer worked with Marco Gritti and Sayamindu
Dasgupta debugged and found the cause of the "Browser being slow after
an update from ship.2 (653) to update.1(690)" issue (Ticket #6046). It
turned out that timestamps of fonts were set in the future. xulrunner
does check if fontconfig is up to date and if it is unsuccessful the
fontconfig is reinitialized and the whole thing repeats itself again.
The resulting loop is causing the slowdown. Sayamindu has provided a
new fontconfig rpm which checks if mtimes are in the future and print
a warning but does return true so Browse will not be slowed down. In
addition Michael Stone has released a new rainbow-0.7.9 which symlinks
'~/.fontconfig to ~/instance'. Simon also provided a patch that
releases exported dbus objects (Ticket #6127), which is important for
activities that run in a single process, such as Browse.

12. Etoys: The Etoys team is working toward delivering a package for
Update.1. Scott Wallace, Bert Freudenberg, and Yoshiki Ohshima fixed
various lingering bugs in the system. Ted Kaehler and Kathleen Harness
are revising the Quick Guide system and contents.  From the effort,
the candidate version for Update.1, Etoys-77.xo, was created. Takashi
Yamamiya and Korakurider keep working on the translation issue.
Takashi discovered that Pootle cannot merge large sets of translations
fast enough.  He is looking into the issue. The interactive geometry
system by Hilaire Fernandez is improved. Now, it is packaged as a .xo
bundle (http://wiki.laptop.org/go/DrGeo). It contains the translation
framework by Korakurider and others for activities written in Etoys.

13. Spreadsheets: Dan Bricklin and Luke Cross are working on a port of
the Sweet SocialCalc Project to the XO laptop. SocialCalc is a highly
functional spreadsheet implemented in JavaScript. (To date, 39
functions have been developed: ABS, ACOS, ASIN, ATAN, ATAN2, AVERAGE,
COS, COUNT, COUNTA, COUNTBLANK, DEGREES, EVEN, EXP, FACT, FALSE, IF,
INT, LN, LOG10, MAX, MIN, MOD, NA, NOW, ODD, PI, POWER, PRODUCT,
RADIANS, SIN, SUM, STDEV, STDEVP, TAN, TODAY, TRUE, TRUNC, VAR, and
VARP.) Manusheel Gupta is helping them with the Sugar port. The first
pass will be to leverage a general application that supports
activities written in JavaScript, with Python-based Sugar binding.

Another approach to developing a spreadsheet activity is to begin from
the GNumeric code base. Manu is working with Jody Goldberg and Eben
Eliason to port a simple version of GNumeric to the Sugar environment.

14. Sensors and learning: Arjun Sarwal incorporated Spanish-language
support into Turtle Art with Sensors and made a slightly modified icon
of the original Turtle Art icon. The activity is now in Joyride Builds
and is identifiable by a slightly modified icon of the Turtle. Arjun
spent time discussing with Edward Baafi of the FabLab and Aaron Miller
of MIT Media Lab's Lifelong Kindergarten regarding how sensors, apart
from acting like an interface to the physical world, could impact the
XO laptop deployment communities. Aaron is working on integrating
sensors into "Scratch" (which is now available for download from
http://wiki.laptop.org/go/Activities); Edward Baafi is interested in
exploring how general-purpose boards, which include sensors as well as
I/O, can be used in conjunction with the XO laptop. Arjun continues to
explore the wonderful possibilities of $2 sensor experiments through
the XO laptop's analog-input port. Documentation of the session with
youth at FabLab Boston is in the wiki (See
http://wiki.laptop.org/go/Measure/Turtle).

15. OLPC Health: The OLPC Health initiative has gained good momentum.
There are active discussions on the Library mailing lists and the wiki
pages have also started to take shape (See
http://wiki.laptop.org/go/Health). A vision document
(http://wiki.laptop.org/go/Health/vision) is also in the works. The
list of group of advisers to the OLPC-Health initiative includes Josh
Hehner, Jim Hopper, Sv Subramanian and Ichiro Kawachi.  More detailed
introductions of the advisers will follow soon on the Library mailing
list. There is a conference call on the 10th of Feb at 1pm EST. People
are invited to propose agenda items by posting on the Health wiki
pages.

16. Ethernet: Michail Bletsas hosted Jonathan Hsu, founder and CEO of
Zoltantech at OLPC this week. He makes a very small, elegant and low
cost (< $10) USB-Ethernet adapter that works well with the XO; it
could be very useful to developers and advanced users.

17. Localization: Arjun, Manu, Bernie, and Walter worked through
integration of patches for Afghan (including Dari, Pashto and Uzbek
variants), Mongolian, Ethiopian, Nepali, and Italian keyboard layouts.
All of them (except Italian and Nepali) are expected to be integrated
into Update. Sayamindu Dasgupta reports that we have new teams for
Italian, Marathi and Sinhala. Continuing on his recent efforts with QA
and testing with respect to local-language support, Sayamindu has
added notes in the wiki on how to utilize the translation testing
features in our web-based translation management system, Pootle
(http://wiki.laptop.org/go/Localization/Testing#Testing_the_PO_files).
He also discovered a few cases where Pootle can become very slow, and
discussed these with the upstream developers. They have suggested a
few solutions; he is trying to implement them in our deployment. The
shifting of Pootle to the new server had caused some issues to crop up
with the Git integration; Sayamindu managed to track them down and fix
them.

Localization of OLPC for Afghanistan: Dr. Habib Khan reports that they
have started localization of Etoys (a major project, as it contains
23,000+ strings). They have done some work in Urdu localization of
Etoys and almost 1000 strings are translated into the Dari language.
The next step is localization of Etoys ino Pashto.

OLPC User Manual: Usman Mansoor "Ansari" and Sohaib Obaidi "Ebtihaj"
continue their efforts; translation of the user manual into the Dari
language is complete and is now under review; the review of the
translated version in Pashto is 90% complete.

Localization Tests: Habib also reports that they have successfully
tested the Urdu localization .po files that Pootle generated on one of
their test machines. There are some small issues with some of the
character bindings; in addition, they are testing it with Dari and
Pashto languages. The target is to have three XOs completely localized
inUrdu, Dari, and Pashto by next week.

18. In the community: On invitation by the Computer Society of
Pakistan (CSP), Habib made a short presentation on OLPC to the
participants and distributed a CD containing presentations on OLPC.
The members of CSP are the leaders of Pakistani IT industry.

Mike Lee reports that the monthly grassroots OLPC Learning Club in
Washington, D.C. had a record 48 attendees (including several children
and teens) this past Thursday night at Greater DC Cares (See
http://www.olpclearningclub.org). Their host, Curtis Cannon, talked
about how DC Cares will use the seven laptops they acquired through a
holiday fundraising effort called Technoliday organized by Peter
Corbett to support their program of pro bono technology consulting for
social change. Justin Thorp demoed the Library of Congress' World
Digital Library, to which he contributed development effort. Mike
demonstrated accessories for the XO including auto adapters, solar
panels, the Weza foot treadle charger, clip on sports viewfinders for
the camera and the new ZoWii miniature USB Ethernet adapter in OLPC
green. Two iLite USB keyboard lights and an auto power adapter were
raffled off. Attendees stayed for another hour to mesh. The first Mass
XO Meet-up was also held this week (in Cambridge).

-walter

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