Deployment meetings/20090120
Jump to navigation
Jump to search
Summary
We learned that:
- m_stone is interested in deployment-related work and that NealS has some cool things going on in Texas.
- dsd is heading to Paraguay on Friday (for 3 months!) with scalability questions, in need of 8.2.1,
- dbennett isn't sure how to transport XOs + solar stuff across Nigeria.
- PY wants to activate 3600 laptops in 10 schools
- dsd/PY would really like to have a build with nandblaster and a fix for #9008.
- (and that PY wouldn't mind working wifi activation either; though currently, they have to touch the laptops to update them anyway, I think.)
- dsd was unfamiliar with the XO-based ivan-act-server code, so that needs to be better documented.
Next,
- we mentioned pia's "deployments wishlist" but didn't discuss it in any detail since pia wasn't around.
- Several of us are thinking about sticking around for the next meeting 11hrs hence)
Finally, cjb and mstone quarreled about the meaning of support and eventually came to consensus that people should act as though OLPC provides no software support; i.e. that they believe the licenses they're shipping the software under. However, cjb expressed some willingness to try to coordinate people who want to undertake mutual support commitments and he hopes that they'll be pleasantly surprised when OLPC occasionally chips in.
Edited Minutes
Intros
m_stone> (We could probably start without Pia, if we need/want to do so...)
m_stone> dsd_: greetings!
CanoeBerry> Yes, feel free to start Michael.
m_stone> okay.
m_stone> Hi.
dsd_> hola
#olpc-deployment vs. #olpc-deployments
sverma> looks like we have two channels going. olpc-deployment and olpc-
deployments
sverma> wiki page has both spellings
dsd_> oops
CanoeBerry> Wiki pages had a typo...
CanoeBerry> Announcing plural instead of singular #olpc-deployment
dsd_> so we move?
sverma> this one has more people
dsd_> lets stay here for now.. we can change next time if needed
CanoeBerry> Wiki paged corrected.
m_stone
m_stone> most folks here probably know me, but, just in case, I'm Michael
Stone.
m_stone> I'm here because I'm interested in seeing XO- and Sugar-based
deployments succeed, because I know a lot about the software and
people such deployments are using, and, frankly, because I'm
looking for work. :)
CanoeBerry> !
m_stone> perhaps others would like ot introduce themselves?
m_stone> unmadindu: hiya! we're just starting introductions.
unmadindu waves
CanoeBerry> unmadindu: feel free to introduce yourself if poss!
dsd_> i think we all know each other?
dsd_> or is this for the logs? haha
NealS
m_stone> NealS: would you mind introducing yourself?
cjb> don't think I know NealS
dsd_> ah true
CanoeBerry> I'm assuming that's Neal from Texas?
NealS> Yes
CanoeBerry> great
m_stone> NealS: we might have met on a s-g call earlier this year?
m_stone> (well, really last year...)
CanoeBerry> Yes, Neal S has been doing his own deployment work in S/E Texas.
NealS> Yes
NealS> I'm still struggling with how I can make my contribution.
dsd_> NealS: could you summarise your work so far?
NealS> I am working on courseware. From discussions with Walter Bender
that is a pressing need.
CanoeBerry> Bryan Berry's opinion precisely.
NealS> Offline I discussed, with Pia, the importance of a focus on meeting
the end-users - the students and teachers - needs. I am hopeful
that this "deployment" group can do that.
A school just SW of London
CanoeBerry> dsd_: aside there a school just SW of London that ordered 10 XO's
CanoeBerry> Small deployment you may want to look into.
dsd_> CanoeBerry: cool thanks
dsd_> g1g1?
CanoeBerry> Yes: RT32112, which means we need to get their permission before
publically discussing. The guy in charge seems really friendly :)
dsd_> CanoeBerry: interesting. that's not a primary school
dsd_> its for 16-18 year olds
Deployment issues
Shall we meet by moonlight?
m_stone> has everyone seen http://wiki.laptop.org/go/Deployment_Wishlist ? m_stone> (those were some requests from Pia about things that I think she wanted to talk about with us) NealS> Hear from Pia? CanoeBerry> N CanoeBerry> Can we get a show of hands as to who's courageous enough to join 3am? m_stone> was thinking about it... CanoeBerry> Ditto NealS> same here sverma> maybe...it will be 12 midnight for me (sfo) unmadindu> I can do that m_stone> CanoeBerry: I'll be dead for the rest of the week; otherwise fine. dsd_> i won't be here CanoeBerry> OK, I don't expect Yama much for the coming days, as he sounds max'd out in Bolivia for a bit :)
Deployment issues redux
m_stone> so... returning to deployment issues?
m_stone> I'd like to see us talk a bit about Pia's wishlist and I'd also
like to get a feeling for "questions that people attending today
would like to have answers for..."
m_stone> I get the feeling that those questions may have more to do with how
best to /contact/ deployments than how to help them, but I still
think it would be good to get the stuff written down.
m_stone> anyone got a better plan?
dsd_> i put my questions on the wiki
m_stone> dsd_: where?
dsd_> http://wiki.laptop.org/go/Deployment_meetings, in the agenda. I
think pia is good to answer q1 because i think she did that in
niue.
m_stone> dsd_: I expect that we'll send an email to all the usual suspects
summarizing our progress on questions. we can highlight your
question there since I think that no-one here has tried it.
dsd_> m_stone: note there are 2 questions, perhaps i should format them
differently? should i give some context on the 2nd?
m_stone> dsd_: yes and yes. :)
Paraguay (PY) scalability questions
dsd_> in paraguay they have government sponsorship for electricity *and*
internet in all 10 schools, which is great and they have money for
school servers for reliable collaboration but I'm concerned because
it seems to me that school server deployments and testbeds are all
small scale so I'm interested if anyone has deployed XS in
multiple-room schools. I cannot imagine what it would be like to
see 100 XO figures on the neighborhood view
m_stone> dsd_: also to be clear -- you're talking about one of Martin's XS
builds? (on some specific hardware?) and you're also worried about
the XOs, I take it?
dsd_> yes, software
m_stone> just about the nbhd view or about other things as well?
dsd_> i dont know about the hardware, but i'm interested in
success/failure stories. well, i'm also concerned that ejabberd
won't be able to cope with so many users?
m_stone> dsd_: martin says that he's squashed the worst scalability bugs
there. (the 1MB-of-SSL-context-per-connection issue in particular).
dsd_> and what scales is he talking about? 100? 500? 1000?
cjb> I think this is what gadget's for. you wouldn't be seeing that many
XO men at once.
m_stone> cjb: dsd is obviously asking for evidence. and for software that he
can use today. so, while gadget may be close on both of those --
it's not here yet.
dsd_> yeah, but gadget requires stuff on both XO and XS side, right? i
don't see that being ready in time for this deployment.
dsd_> m_stone: evidence, experiences, whatever. but yes, i'm interested
in actual field usage stories. i know XS is still quite new, so i'm
not sure how easily we'll find those but again i wonder if pia
would know.. sounds like it might be a similar situation in niue.
martin's input would also be welcome :)
m_stone> dsd_: I just read through the last 20 or so emails martin sent and
he doesn't mention any concrete test results.
Paraguy (PY) scheduling questions
m_stone> when do you need numbers by?
dsd_> not sure; i'm not in paraguay yet so i don't know the exact
timeline and i don't know what i'll be working on first, second,
etc.
m_stone> okay. when are you arriving in paraguay?
dsd_> friday, staying for 3 months
CanoeBerry> dsd_: 3 months is amazing. any compadres possibly joining you to
help, eg. from the rich west here or otherwise?
dsd_> CanoeBerry: mm.. i almost persuaded brian, but he needs funding
which they are still working on
dsd_> there is a diary of the problems encountered during py teacher
training here: http://wiki.paraguayeduca.org/index.php/Diario_Sugar
en español..
dsd_> we also really want 8.2.1 for paraguay, for the mouse fix (#9008)
(and maybe nandblaster). they saw the recalibration failure a lot
during teacher training (300 XOs)
cjb> turning the crank on 8.2.1 is my top priority at the moment.
still, it will probably take a week or three to hit final. most of
the legwork is done, though.
dsd_> i proposed (and will follow up on) that i work on 8.2.1 as an
immediate priority when i am in paraguay so maybe i can help you
keep things moving...
cjb> dsd_: oh, that would be a great help; thanks!
unmadindu> cjb: let me know if you need any assistance for 8.2.1
cjb> thanks all :)
m_stone> dsd_, cjb: #9008 is 2-months stale. anyone know what the state of
the patch is?
dsd_> m_stone: its included in staging-8 or staging-9, i think
m_stone> okay. have you gotten a positive test result for it? (i.e. has
somebody who has seen this problem tried the new kernel and said
"why yes, I can no longer reproduce the issue....")
dsd_> i dont think so actually... i might be wrong though, its been a
while
m_stone> cjb: so it sounds like there's a little bit of work necessary there
on both the positive (does it fix something?) and the negative
(does it introduce any regressions?) sides...
m_stone> cjb: can you comment on what level of QA you're trying to provide
for 8.2.1? (that's a royal "you" by the way)...
cjb> m_stone: not sure yet. Mel hopes to continue to lead the testing
team for 8.2.1 testing and we will probably also forcefully
encourage Uruguay and Peru to offer feedback :)
Paraguay (PY) security goals?
m_stone> dsd_: also, what is the security status of paraguay's laptops? how
physically distributed are they? (i.e. if you had to send somebody
out to re-nand-blast them, how bad would that be?)
dsd_> all 3600 will all be within 1 city distributed among 10 schools.
they requested them unactivated and they want to activate them in
schools. they were asking me about activation-via-school-server and
stuff...
cjb> connectivity?
dsd_> internet in every school
dsd_> i think i could probably hack up an XO to distribute the activation
keys over the mesh.. that would save us a lot of hassle, but it
isn't really sustainable for scaling up the deployment
m_stone> dsd_: no hacking needed; already done. see
http://teach.laptop.org/~mstone/ivan-act-server.tar.gz. well-tested
in UY.
dsd_> m_stone: is that easy to setup?
m_stone> dsd_: trivial. you just unpack it and run ./install or so. it takes
a big blob of leases in the usual json-dict format.
dsd_> great
Meanings of Support?
cjb> dsd_: we're going to have to have a sobering conversation at some
point, where we tell them that no future work is scheduled on
security features, and we will be actively deprecating our support
of the current systems
m_stone> cjb: uh oh. yet another deployment actually seems to care a little
bit about our little "a"rm scheme) [it's really analog rights
which are being managed...]
cjb> (which means that if they want to offer that security, they should
grab the source and setup all their own systems to provide it)
cjb> dsd_: but we're not ready to have the conversation just yet, just
warning you
dsd_> good news is that the paraguay people are really great so i think
they will understand that they have to figure out a way to work
through these "hard times"
cjb> m_stone: countries would like us to do lots of things for them.
how much they want it is unlikely to change our mind about whether
to provide it going forward, though. besides, we're not saying they
can't do antitheft, we're just telling them that they have to
support it themselves...
cjb> ...which was a pretty reasonable thing to have been aiming for from
the start, in my opinion
m_stone> cjb: just like we're not saying they can't install their own
builds, just that we won't help them when they do. same effect,
imnsho.
cjb> m_stone: in that both consist of other people wanting us to perform
self-destructive and unsustainable acts for them, they are similar
m_stone> cjb: what you ought to be doing is saying that you know how
important feature ____ is to them and that you'd like to help
coordinate the people who want feature ____ so that they can share
the costs.
cjb> m_stone: yes, that is probably what we will say.
m_stone> cjb: I never expected to hear you call exercising the Four Freedoms
"self-destructive and unsustainable".
cjb> um, you didn't. are you trolling?
m_stone> cjb: no, I'm searching desperately for a sound position from you
and nearly crying because I can't find one. (in frustration)
dsd_> he did say that he's not ready to have this discussion yet
cjb> m_stone: my position is merely that deployments will have to start
bearing the burden of antitheft and build support themselves,
because we can't afford to do it anymore. I don't mind how they
choose to organize themselves to do this and I'm happy to help in
the organizing, if that's the best way I can help. I don't think
I've said anything to the contrary of this, though miscommunication
is entirely possible.
m_stone> cjb: I reacted most strongly to your phrase "...and we will be
actively deprecating support for...". the reason being the meaning
of OLPC's "support" to date which is essentially that if OLPC isn't
supporting it, then it shouldn't be done.
cjb> deprecating *our* support for, yes. the reason our "support" has
been so powerful in the past is precisely because we were doing
what I am now advocating we stop doing.
m_stone> cjb: so is it fair to say that we should act as though OLPC
provides no software support whatsoever? i.e. as though all the
software is provided under the usual "no warranty of fitness for
any purpose..." and that's exactly what is meant?
cjb> m_stone: I think that would be a worthwhile assumption to hold, and
may provide for pleasant surprises when we do offer gratis support.
m_stone> cjb: I'm glad we're agreed.
cjl> m_stone: welcome to the crowd outside, we get fewer hints about
what OLPC thinking is, but the smart ones amongst us try to care
less about it and don't get even the maddening tidbits the insiders
get to make us crazy :-)
m_stone> cjl: :)
CanoeBerry> PS Reuben isn't going anywhere as far as I know, doing basic SW
support :)
dbennett's questions
dbennett_> I have a question. It's about supply chain, or logistics. I'm
working on a small pilot in Eastern Nigeria, and we're planning on
transporting large, heavy, high-value goods across half the country
via truck. My guess is that the issues inherent in this
transportation problem are in some ways location specific, but I'm
having trouble wrapping my head around the problem of finding a
trustworthy vendor to carry out this work. How do we select someone
to move 100 xo's, and a significant quantity of Photo Voltaic
equipment across the country?
CanoeBerry> airport to where?
m_stone> well... do we know anyone else who has already dealt with this sort
of problem who we could ask for advice?
dbennett_> I'm planning on relying on my local contacts to help solve this
problem, but most of them have never tried to do anything like this
before.
m_stone> dbennett_: if the solution is in their hands... what are their
concerns?
dbennett_> m_stone: I haven't really turned the problem over to them yet.
It's still on my to do list. Just fleshing it out here is giving
me some new ideas though. I think I'll try some of the business
contacts we've developed instead of using the Church/Missionary
contacts. They may be much better connected in the capital city
that will be receiving our goods from abroad.
CanoeBerry> dbennett_: just to be clear, i'm interested in Nigeria
geographic/logisitc details when you have a chance, but you may not
want all that published..
dbennett_> CanoeBerry: I'll try to keep you posted on any progress. Don't be
afraid to hound me a bit though if you're not hearing anything.
cjl> dbennett_: do you have any Peace Corps contacts tha may have faced
a similar problem?
dbennett_> cjl: Nope, no Corps contacts in Nigeria
cjl> dbennett_: I have a few links on a sandbox page of NGOs that help
other NGOs http://wiki.laptop.org/go/User:Cjl/Sandbox4 never know
if there might be a useful link to follow on those sites.
dbennett_> cjl: Thanks, I'll take a look.
CanoeBerry> cjl: impressive list, we need to talk :)
cjl> CanoeBerry: sdteal away, it's a wiki.
cjl> It's slightly stale and I'm sure there are more. .