[Community-news] OLPC News (2008-04-26) Technology Team
kim at laptop.org
Sat Apr 26 11:06:06 EDT 2008
* Tech Team in Deployments:*
In Montevideo two weeks ago, team Ceibal and Martin Langhoff worked together
for two days. Focus was on mapping a road to convergence with the NOC that
looks after School servers, and better communication with 1CC. The Ceibal
team has significant depth of experience with XOs in the field, their
lessons learned and effective scripts are gold.
In Lima more than a week ago, Martin worked with the local team, fleshing
out AP testing strategies, XS roadmap, and XS administration infrastructure.
John Watlington, in Peru this week, has been in discussion about school
server specifications. It became clear once again that cheap, and
environmentally robust school server hardware is a missing part of the OLPC
system. Peru's previous experience with using traditional PC hardware in
the jungle and the Andes was that the machines failed frequently, due to the
heat. Listing the required environmental
requirements in the specifications results in a multiplication of the price.
Complicating the process is the fact that most bidders are assembling the
units from whatever components they have on hand, resulting in varying
quality across the entire lot.
Martin Langhoff reports lots of hacking around xs-config. He shaved all the
yaks necessary to get portable xs-build actually building live CDs
successfully. PyAr programmers Alejandro Cura, Lucio Torre and Ricardo
Quesada worked with Martin and held 2 XS-focused Sprints, working on idmgr
(security and registration lag) and CDpedia (yet another wikipedia slicer).
(2 weeks ago, catching up on weekend reports)
With Tomeu Visozo's help, XO-to-XS backups are finally moving forward. The
XS image is seeing some progress in minor installation bugs.
A build of the Multi-Battery Charger of 50 units is in the planning stages.
The final few tweaks to the tooling are happening this week. It currently
looks like this will happen the 2nd or 3 week in May.
Richard Smith, Scott Ananian, Andres Salomon, & Chris Ball discussed some
possibilities for enhanced power management. Ideas are brewing, more
structured planning should occur soon. Andres and Richard began to lay out
the plan of attack for achieving the sub-200ms resume goal. Work will begin
as soon as Andres is finished with his upstream kernel merge.
Chris Ball began thinking about a user-selected low-power mode for the XO,
to showcase our power management work, Trac #6935. This would be an "under
the covers" mode where we prolong your battery life by any means necessary,
including disabling networking (which is currently the major obstacle to
pervasive power management all the time).
Eben Eliason cleaned up and pushed Journal redesigns, now in joyride, and
worked with Martin Dengler on improving/adding devices to the Frame.
Scott Ananian spent some time this week assessing the technical viability of
Sugar on Windows. Please look for his Mini-conference video postings at
http://wiki.laptop.org/go/Apr_3-4_Mini-conference. Also, he started
gathering useful content from his FISL trip at
Chris Ball resurrected our Tinderbox tool, and told it all about the
joyride, faster and update.1 build streams as well as how to build
sugar-jhbuild regularly. Michael Stone began to outline some documentation
for how volunteers can help us by running and improving it themselves at
Michael Stone prepared rainbow patches allowing faster activity launching
and sugarization. He submitted them for review to code review. He also
reviewed backup bugs, then organized a software status meeting on datastore
and backup bugs.
There was much discussion this week on the mailing lists and in forums
regarding OLPC's commitment to open source, transparency, and how to improve
communications. Most of the team spent time working with the community, in
discussion and in writing their own thoughts on how to move forward. There
will be more discussions and a meeting with Nicholas next week when he is
back in town.
Sayamindu Dasgupta took the existing Pootle user's Guide presentation
created by Sameer Verma and updated it to reflect the new look and feel of
our Pootle installation. He also added a few more slides, highlighting some
of the special cases one might encounter while working on the translations,
as well as some tricks that might make the translation easier. The slides
are available at
Simon Schampijer has been doing stabilizing work by cleaning the sugar code
Guillaume Desmottes ressurected the video-chat-activity and packaged the
needed bits to make it work with current Joyride. It's far from
being perfect but it should be usable for simple tests (he successfully
established audio/video calls between 2 XO's and between 1 XO and
Empathy). See http://lists.laptop.org/pipermail/devel/2008-April/013227.htmlfor
It appeared than some collaboration problems are due to ejabberd bugs. Need
to contact Process-1 guys. Maybe we could add a "ejabberd" component to trac
assigned to them?
Morgan Collett tested ejabberd installation instructions (
http://wiki.laptop.org/go/Installing_ejabberd) with Ubuntu 8.04. He also
worked on Sugar in Ubuntu 8.04. He extended the HelloMesh collaboration
example to send a simple text control over the Tube, with alerts to show
what is happening. He fixed a bundlebuilder problem in Sugar introduced by
the pylint patches and worked on Chat UI. Please send Morgan any
collaboration-related news, or add it to
Dafydd Harries worked on introductory documentation to Telepathy framework,
helped Guillaume get started on working on Gadget, and began mentoring Assim
in his Summer of Code project.
Dennis Gilmore reports that we can now submit activities for review. Fedora
9 has working support in GDM/KDM to log into sugar. Draft guidelines can be
will be moved to the final location very soon. This will offer a great
platform for developing activities. it will also help sugar be more
accessible to people. Please package and submit your activity. Please ask
Dennis if you have questions.
John Watlington, in Peru this week, performed tests against different access
points, while the Collaboration and Mesh Testbed in Peabody was dismantled
and is in search for its new home. Two more low-cost access points (one from
Linksys and one from DLink) have been tested against a classroom of fifty
XOs, and seemed to perform fine.
Ricardo Carrano reports that tests with firmware 22.p10 during the week
revealed that we seen to be free of #6589, but we still have issues with the
multicast filter (new issues were found in the proposed path and addressed
during the week). With a test kernel, we can now run firmware 22.p10 without
problems in nodes were idle suspend is not active, but a suspended node
still does not react as expected (it is incapable of detecting a new
neighbor or shared activity). The next step is determining if there are
issues in the firmware or if the driver patch is still not satisfactory.
He is attempting to build a clear view of current wireless issues and
started a wiki page which focus on problems to be addressed in the libertas
firmware or driver but which also lists other network-related issues like
middleware, hardware, school server or applications (
http://wiki.laptop.org/go/Wireless_Issues_Apr08) On the same note, I am
trying to update the open tickets and close the fixed ones. Contributions to
the page and to the tickets are more than welcome.
Richard Smith helped out with the OLPC manufacturing server some more this
week. Turns out some of the issues were not as bad as originally thought
and we were able to work out most of the issues with the erroneous data we
got from Quanta. They claim to have added some checks that will help
prevent future occurrences.
Richard Smith wrote to the community to address issues with keyboard
failures due to sticky keys. The manufacturers of the keyboard have worked
with OLPC to make changes to the internal construction of the keyboard. We
strongly believe that these changes should fix the sticky keys issue. These
changes have been rolled in over the last few weeks of production.
If indeed the problem is much more widespread then OLPC needs more accurate
data on the failure rate. OLPC invites the community to help.
If you currently have an XO with a sticky key then please add your serial
number to the following wiki page and what key(s) appear to be
Emily Smith reports that she and Adam Holt have successfully created a
shipping order to Brightstar for 122 replacement laptops, and for spare
parts for our first potential repair centers. No replacements had been going
out in the last several weeks due to problems ranging from oversight of the
RMA program to detailed format problems with our ship orders. Communicating
shipping orders with BStar has been a problem for weeks and donors continue
to be frustrated with these delays. We hope we have the ship order detail
figured out by early next week with much help from Christine Myrick and
Emily spoke with about 5 donors on the phone regarding G1G1 refunds and
Getting Started questions. The majority of her time is spent answering
emails regarding G1G1 questions ("where's my laptop?" is most common,
followed by "how do I use this thing?" and "It's too difficult to use, I
want my money back!"). She also sends out tax receipts for people who have
re-donated their laptops back to us (about 70 in that queue).
Adam Holt reports that he continues to work on tons of escalated shipping
and billing tickets. He is very hopeful that the deluge will ease by around
May 1, *IF* Patriot's Donor Services continues helping. He and Sandy Culver
have been surveying Donor Services g1g1service at yahoo.com and
xogiving at gmail.com and believe that these are getting about 15 times more
mail than our RT site.
Adam is working with 15 worldwide community repair centers' and sent out the
first broken laptops to them to seed their centers. He is also collecting
updates from the Support Gang for next week's G1G1 PostMortem, lessons
learned, so that the extensive experience can be used in future programs.
Adam and Kim Quirk discussed a 'Saner Future', so that truly progressive
groundwork can be laid out in the Support group starting in May and June.
Henry Edward Hardy placed the order for three new servers to replace
pedal,crank and xs-dev. We expect to receive these within a few weeks and
start configuring them. Thanks to everyone who helped with the specification
and ordering process!
We experienced significant downtime at 1cc this week due to an unscheduled
electric power interruption. Everything at 1cc was down for several hours.
Public facing services on crank and pedal such as git repositories, mailing
lists and mail aliases, OLPC website and wikis, were apparently unaffected.
We are done with testing and pre-configuring the Big Sister monitoring
service. This will assist in monitoring the health of our systems and give
us some metrics for uptime and service accessibility and response times. We
plan to deploy the Big Sister server on the xs-dev replacement.
There has been a power-related crash which has damaged the filestructure on
the Chinese-deployed OLPC manufacturing data server. Richard Smith and Henry
are assisting in the recovery efforts.
Charles Kane has produced a wonderful Internet slideshow which is posted on
YouTube at http://www.youtube.com/watch?v=s4ojFcZIqRU. He was assisted by
Michail Bletsas, Darah Tappitake, Jennifer Amaya, Henry Edward Hardy and
Bernie Innocenti of OLPC-Europe.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Community-news