[Community-news] OLPC News 2007-10-20

Walter Bender walter.bender at gmail.com
Sat Oct 20 14:49:47 EDT 2007


1. Lima: Carla Gomez Monroy's hard work at Arahuay—the town which
hosted the first OLPC pilot in Perú—came to fruition with news that
the Budget Commission of the Peruvian Congress has unanimously
approved of the first allocation of laptops for children. The
announcement was made just as Jim Gettys, Hernán Pachas, and Carla
Palomino Guerrero arrived in Arahuay for a visit. Hernán and Carla are
working with Oscar Becerra, General Director of Educational
Technology, Ministry of Education of Peru, who is in charge of the
OLPC program in Peru. (Jim gave the keynote address at the Vision 2007
Conference in Lima, Peru, having worked with Hernán and Carla earlier
in the week.)

2. Miami: Nicholas spoke at the Latin American Society for Information
and Publishers, which drew all the major news media from Latin
America. The audience carried stories throughout Latin America the
rest of the week, widely naming Uruguay the first country to boldly
proceed.

3. Accra: Matt Keller, Juliano Bittencourt (who is one of the leads at
the Porto Alegre OLPC trial school), and David Cavallo ran a Laptops
and Learning workshop as part of the build up towards Ghana's
deployment of One Laptop per Child. The workshop was held at the Kofi
Annan Center and was organized by the Ministry of Finance, who is
championing the initiative. Participants came from a broad array of
Ghanaian society, including various ministries, universities, NGOs,
and the private sector. The goals of the workshop were to familiarize
the participants with the ideas behind OLPC; to learn about learning
and laptops; to learn about experiences in other countries to date;
and to plan next steps towards a broad deployment.

4. Springfield, Illinois: Cavallo gave a keynote address to the Rural
Telecom Conference in Springfield. There were participants from many
states as well as tribal reservations; there was tremendous enthusiasm
for deployment of OLPC in US rural areas and twinning them with
deployments in other countries.

5. Cambridge: SES-Americom has given us a two-way satellite terminal.
The installation was completed this week at the roof of the Media Lab
and the terminal is now connected to the OLPC offices via a Motorola
canopy base station that sits on the top of MIT's Eastgate building.
In this manner we are now able to setup a simulation of a complete
rural school network at the OLPC Cambridge office and test network
connectivity via satellite. The system that SES has given us is an
iDirect series 5000, which is towards the high-end of the spectrum as
far as VSAT routers go and is able to sustain 500Kbps symmetric with
download bursts up to 6Mbps.

6. Laptop on your lap: XO is the first "laptop" that can safely be
used on one's lap. Most laptops are officially called notebook
computers, because use on a lap is unsafe—you risk burning your skin.
Not so with the XO; it is so low power that the electronics don't get
hot; and further, the electronics are next to the screen, not the lap.
The plastic in the keyboard area stays cool all the time.

7. Mesh: Ronak Chokshi from Marvell and Javier Cardona from Cozybit
made an "emergency" visit to OLPC. The visit resulted in some useful
discussions that brought everybody up to speed with the status and the
development of XO's wireless stack. Marvell released firmware
5.110.19.p0 on Thursday; we reached consensus that we need to
incorporated into our builds as soon as possible for testing.

8. Wireless driver: The effort to push the wireless driver upstream
has left the driver on our builds with a limited "control knobs." Of
the 79 documented iwpriv controls, only 28 are exposed in the current
builds. Of the 51 that don't work, some are rarely used but some are
essential, e.g., radioff. Given that many of the enhancements that we
have requested Marvell to implement depend on driver support, in the
interim, Michail Bletsas recommends that we use the original
libertas-2.6 tree for our builds—where Marvell and Cozybit are
contributing—not the tree that we are using for pushing patches
upstream.

9. New intern: Ricardo Carrano from Universidade Federal Fluminense
(UFF) in Niterói, Brazil has joined OLPC as an intern for the next two
months. He will be working on various wireless issues with Michail.
Ricardo has been instrumental in performing a variety of tests since
the beginning of the year in Brazil.

10. Bulk charger: The industrial design for the multi-battery charger
has been approved. A full mechanicals package has been created and is
off to tooling vendors to get some quotes. Gecko had a
stereolithographic (SLA) model of the front panel build with five
battery slots in order to test the battery retention system, which is
designed to allow battery removal with one hand. We should have a
mechanical model in Cambridge soon for evaluation. Flextronics
finished up the thermal simulation of the inside of the charger.  We
are looking at a 15-degree C rise over ambient temperature.
Simulations are showing that we are well within the operating range of
the components while operating at our target of 50-degree C ambient
temperature.

Bitworks and Richard Smith brought up the CPU board for the bulk
charger: Forth is running and all the I/O paths have been exercised.
Both Richard and Lilian Walter have prototype CPU boards for software
development. The first draft of the software specification along with
the relevant EC code that currently runs as part of the laptop
charging system were delivered to Lilian last week. She has ported the
USB code and has written a software pulse-width modulator (PWM), which
will be used to drive the charge voltage for each battery. She is now
working on increasing the PWM interrupt priority. Next week Richard
hopes to charge some batteries.

11. Cow power: In a related effort, Richard has been working with
Arjun Sarwal, helping him refine his plan for charging the XOs using
cow power.

12. Firmware: Mitch Bradley has been putting together a firmware build
for mass production. Q2D01 was officially released, but we may decide
to bump to Q2D02 to address some issue with security migration.
Q2D01 has several important bug fixes: wireless association, JFFS2,
and support for non-US keyboard layouts, as well as a secure OS image
update.

13. Embedded controller: An agreement between EnE and OLPC on use of
the EC source code has been finalized—now it just needs signatures.
Richard, Mitch, and Scott Ananian worked up a method for automatic
firmware updates in the face of our full security mode. It needs an EC
patch that might not make it into MP. In other news, a small "buglet"
with the Read Power Rail command (0x2B)—it always reported the DCON as
powered up—has been fixed.

14. Laptop hardware: John Watlington spent the week working on
improving our test setup and chasing the occasional "reboot on resume
on ping" bug. This is our most significant remaining hardware
problem—it occurs rarely enough that, while we will hunt it down and
fix it, we will not stop initial production. An update will be
provided next week, after John and Chris Ball have tested a
statistically more significant number of C2 pre-build machines at the
factory in China. We are releasing the modified B4 and C1 laptops from
the Cambridge test bed to software developers (four had already been
sent out). Kim Quirk is distributing these. Two will be kept running
in the testbed in a continually ongoing attempt to break the
one-million suspend/resume-cycle mark.

We have gotten to the point in fixing the suspend/resume problems
where we are once again finding subtle problems with our test setup.
Javier and Ronak helped us debug some remaining wireless-mesh
problems: we started testing with a new build of the WLAN firmware and
consequently started seeing more instances of the problem loosely
identified by Trac ticket #1752 (USB wireless suspend/resume failure
at setup phase). This is the ticket that gets blamed for problems with
networking after a large number of suspend/resume cycles (over time, a
collection of bugs have been corrected by multiple changes in WLAN
firmware, EC firmware, and the motherboard itself). We hadn't seen the
bug in modified machines; we will be testing an even newer WLAN
firmware, which might correct the problem.

15.  Javier and Ronak also looked at some problems we have been having
with the active antenna (it occasionally stops working, even though to
system software it appears to be OK). As this problem only happens
after weeks of operation, they weren't lucky enough to see the
behavior.   What they did realize was that the test setup Marvell had
been using in India was completely different from the one in which we
were seeing the problem. They will now run the school-server software
in their testbed. Again, Marvell's latest firmware might fix this
problem.

16. Peripherals ideas: Joshua Seal has created a new discussion space
for peripheral ideas. It combines all previous ideas posted on the
OLPC wiki regarding power, connectivity, sensors, and input and output
devices. By going to
http://laptop.org/teamwiki/index.php/Team:Peripherals you will see all
of the peripheral ideas and are welcome to post any new ideas you may
have. Alternatively please email your ideas to josh<at>laptop<dot>org
with your suggestions.

17. Sugar: Simon Schampijer worked on Sugar refinements to the
standard alerts and added a timeout alert. He also added full-screen
support to Sugar—hides the toolbar and tray—and add tray support in
activity windows. This is implemented and works for all the Python
activities, since we handle this in the base activity window. (He has
also added support of trays to the base activity, which is why both
the toolbar and the tray are affected by full-screen mode.) The Browse
activity is a good place to see and test this new feature. Alt+enter
is bound to this feature. (Note that currently you have to use
Alt+shift+enter due to a misbehavior in the Matchbox window manager.
This has been filed and will hopefully be fixed upstream early next
week.) Simon also added alerts for "Download started" and "Download
completed" and feedback for downloads in progress when closing a
session.

Internationalization support has been integrated into Measure Activity
and with help from volunteers Spanish and Portuguese translations have
already been done.  Arjun Sarwal has been working on displaying
multiple logs at the same time to enable comparison amongst logs.

Bert Freudenberg spear-headed the effort to adapt Squeak to the API
changes and security container change after Trial-3. Takashi Yamamiya
also take the charge of DnD fixes. Yoshiki Ohshima worked on the fixes
on accented character input and support for the DC-mode input. Scott
Wallace gathers numerous fixes from community and merge them to the
Etoys.

Hilaire Fernandes has led an effort to port Dr. Geo, a GTK interactive
geometry software to the laptop. It allows one to create geometric
figure plus the interactive manipulation of such figure in respect
with their geometric constraints. It runs inside of Etoys; a great
example of "hard fun." (See http://wiki.laptop.org/go/DrGeo for
details.)

Muriel Godoi continues to make improvements to the Memorize Game
activity. He has reduced the memory footprint; added new icons to the
"create" mode (which lets children create their own games); integrated
the datastore; enabled importing of pictures and audio from the
Journal (for building cards for new games); added an icon and tool tip
for a "reset game" button; and incorporated the new protocol for
sending files over tubes in order to share a child-created game. A new
release will be available soon.

Manusheel Gupta has been working with the team from LSI at the
University of São Paulo (including Alexandre Martinazzo, Irene
Karaguilla Ficheman, Nathalia Sautchuk Patrício, Roseli de Deus Lopes,
André Mossinato, Rafael Barbolo Lopes, and Pedro Kayatt) on the Paint
activity. They have some new ideas for implementing sharing and fixing
bugs. One important enhancement is Journal integration for image
imports. (Try downloading
www.lsi.usp.br/nate/documentation/Paint-12.xo).

18. Keyboards: Manusheel, Sayamindu Dasgupta, Roshan Kamat, Tushar
Sayankar, Jens Peterson, and Walter Bender have finished the layout
for a Deva keyboard (See
http://wiki.laptop.org/go/Devanagari_Keyboard). We hope to finish the
Nepali and Pashto keyboards in the coming weeks. Manu is leading a
discussion on an OLPC keyboard for the blind. Please send your
ideas/feedback to manu<at>laptop<dot>org).

19. Schedules/testing: We have done what we hope is the final OFW and
Trial-3 builds for the mass production start at Quanta. The focus of
the development and test teams starting this week has been mostly on
first deployment (FRS) bug fixing and testing. Yani Galanis, Ricardo
Carrno joined Ronak and Javier to help sort out issues with wireless
connectivity. Alex Latham continued to help out on suspend/resume
testing as well as updating and executing test plans from the wiki
test area (http://wiki.laptop.org/go/User_Stories_Testing).

20. Test sprint: Michael Stone ran a "Test" sprint at OLPC to
brainstorm ideas for improving our QA and test process. Thanks to
community members who contributed (on IRC) including Titus Brown and
Grig Gheorghiu (two testing/QA experts from the Python community),
Mitchell Charity, Ricardo Carrno, Chris Ball, Danny Clark, Reynaldo
Verdejo, Marco Gritti, Alex Latham, Yani Galanis, and others.
Accomplishments include: specific work towards an improved tinderbox;
the creation of a vision for the test activity; an attack of the
problem of uploading a log set from the laptop to a central server;
some updating of our current set of test plans; and discussions on the
Trac extension for test-case management. We will follow up over the
next few weeks (See wiki.laptop.org/go/Test_process_sprint). Weekly
test meetings are held on Mondays at 1pm EDT. Please send email to
kim<at>laptop<dot>org if you want to be on the testing mail list.

21. System updates: Chris Ball modified the tinderbox to run on
Joyride builds and to launch every installed activity; recording how
long each launch took. Chris is also working on having open hardware
manager (OHM) interpret the wakeup reason newly reported by the
kernel: if the wakeup reason suggests it, we will go immediately to
sleep after waking up. We can also start treating "suspend"
differently from "sleep"—if the power button is pressed, the
assumption is that user wants us to stay asleep until we get another
power-button press to wake us up.

This week Andres Salomon worked on bugs, including the infamous JFFS2
corruption bug that appears to be a race in some VServer or VFS code
and has proven difficult to grasp. Another bug was the inclusion of
the DCON smbus init recipe, which (we believe) makes the DCON finally
behave as we intended. Andres also updated the kernel's geode GPIO API
and wrote a patch that allows you to see what event has woken the
machine up via /sys/power/wakeup-source. He sent more patches upstream
and helped massage outstanding patches into 2.6.24-rc1. Andres merged
a few of Mitch Bradley's patches, as well: power off via cs5536 rather
than the EC (which allows us to drop indexed I/O from the EC, unfreeze
the DCON when (not if) the kernel panics, etc. SELinux was disabled in
the kernel; it was originally enabled so that people could use it as
the starting point to come up with something comparable to Vserver. No
one ever did and one of the bugs Andres worked on (that appears to be
caused by hardware memory corruption) has SELinux stuff in the
backtrace, so it was time to remove it.

22. Security: Ivan Krstić worked on the system of capturing
manufacturing information from the production line for the
security/theft system—a server will be set up at the factory. As a
fall-back option, we will utilize a parallel posting of manufacturing
data directly upstream to Cambridge. We have requested the source code
to Quanta's manufacturing data submitter program to verify it can
handle error conditions correctly. Ivan also spent several hours with
Mitch Bradley debugging a kernel filename corruption bug triggered by
VServer's copy-on-write functionality; we didn't succeed in
specifically tracking down the bug, but we identified locking problems
in a VServer routine that are being fixed.

23. Backups: Ivan is working with Marco Gritti and Tomeu Vizoso on the
school server components needed to provide a human-readable index of
Journal. Ivan wrote a detailed specification and has iterated with
Marco and Tomeu, who will be working on the "XO side" of backup, to
meet all the requirements (See
http://wiki.laptop.org/go/XS_backup_restore).

24. Activation/updates: Michael Stone spend time this week making some
architectural changes to the mechanism we use for selecting packages
to be built by pilgrim; Scott Ananian implemented part of one of the
requests, namely, to require that detailed change-entries be supplied
describing what each package in a fragment of a build-branch is
intended to accomplish.

Mitch Bradley, Kim Quirk, and ScottI tried to nail down the bits that
are going into the MP machines. We will shortly have a Q2D02 firmware
and Build 618, which we hope is ready for MP. Many of Scott's fixes
for early-boot code landed in Build 617, including better
activation-failure messages—more translations needed—and activation
from a school server.

Scott set up updates.laptop.org on its own (virtual) machine; he built
 and installed activation.laptop.org, which allows deployment teams to
generate activation leases and developer keys. He also installed a
"recent" build on all 25 meshtest machines hanging from the ceiling of
the OLPC office, reclaiming some machines that had left the mesh for
other testing.  There is now an infrastructure in place for doing
automatic upgrades of the entire meshtest, which will let us easily
put tests in place and execute them, as well as load testing our
upgrade mechanism.

25. Text to speech: Hemant Goyal and Asiem Deodia are working on
getting a text-to-speech synthesizer integrated into xbook (and
perhaps into Sugar). Arjun Sarwal is mentoring them as part of Summer
of Content. They have an initial design working, involving a dbus
service that will capture highlighted text and play it with espeak.  A
global "play" button is planned for the Sugar activity toolbar so that
this is accessible from all activities. Paulo Condado, a programmer
from Portugal introduced by Antonio Battro, works on accessibility
software for the disabled. He is working to make a version of his
"easyvoice" tool work on the XO; he is also beginning with
text-to-speech services (See
http://w3.ualg.pt/~pcondado/easyvoice/).

26. How-tos/documentation: A few groups have independently developed
their own "how-tos" about using Sugar and the XOs. Christoph Derndofer
and Eduardo Silva each took a stab at how-tos for using activities and
Todd Kelsey and Val Scarlatta worked on updating the 542 Demo Notes
with more detailed information from the wiki and updates for recent
builds. John Gilmore wrote in with his own ideas for help files. There
is a group discussion planned for next Saturday to bring these similar
works together.

27. Localization: Our pootle instance is running, thanks to Xavi
Alvarez and Rafael Ortiz, and will be used this weekend, though
integration between pootle and git has been harder than expected; much
is not yet automated.

28. Design:  Aza Raskin of the Humanized design team and David Huynh
of CSAIL's Simile project are coming to OLPC late next week to discuss
some of their design ideas and to see how their teams can help with
interfaces to display content and personal history.

29. Community: Manusheel Gupta and Arjun Sarwal finished work on a
community and partners database for India, to be integrated with the
OLPC wikis and made available to others.

30. Copyright: Pam Samuelson from Berkeley Law School has offered to
help evaluate copyright concerns for media and content we distribute.
A meeting is being scheduled for the end of the month at Harvard Law
School.

31. SimCity:  Don Hopkins is done with most of the SimCity port, but
is still working on sound and font issues.  He is working on the
project again this coming week, with help from Julius Lucks.

32. Creators/curators: Roberto Faga has started working with Marcelo
Bursztein and the ePals team to finish their simple activity for
school-to-school collaboration. Elizabeth Stark and the Free Culture
Foundation have found a PR group in New York interested in promoting
"music for OLPC." (This needs to be coordinated with W2.) They are
adding one or two new collections a week to the "olpc" category at
Jamendo. (www.jamendo.com/olpc should be up sometime this week with
the OLPC music lists.) Colingo has new Spanish- and
Portuguese-language videos out (See
http://wiki.laptop.org/go/ColingoXO). Wikihow and howtopedia are
working together to coordinate their efforts in English, Spanish, and
French. The OLiVER project, working on publishing curricula across
Africa, is working with wikieducator to share their materials.

33. Java: Adam Bouhenguel has started working on a CDC port to the XO.
Stefano Mazzochi has a plan for how to cut down the Harmony
implementation for comparison, and is trying this out in emulation.

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


More information about the Community-news mailing list