Web   ·   Wiki   ·   Activities   ·   Blog   ·   Lists   ·   Chat   ·   Meeting   ·   Bugs   ·   Git   ·   Translate   ·   Archive   ·   People   ·   Donate

#sugar-meeting meeting, 2010-12-18 14:04:53

Minutes | Index | Today     Channels | Search | Join

All times shown according to UTC.

Time Nick Message
14:04 meeting Meeting started Sat Dec 18 14:04:53 2010 UTC. The chair is tch. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:04 Useful Commands: #action #agreed #help #info #idea #link #topic #endmeeting
14:04 tch its alive!
14:05 #topic Silbe's update on patchwork
14:05 silbe tch: you can use "/msg meeting commands" to see the first few commands. I haven't figured out yet how to get a full list from the bot itself, but there's documentation in the Debian wiki.
14:08 tch i think already defined the topic haha
14:08 guess you should just start?
14:09 silbe Patchwork should work well enough AFAICT. Re. the review process, erikos and I will step up as maintainers for sugar, sugar-base, sugar-presence-service and hulahop. I'll announce it later today on sugar-devel.
14:10 The proposed process is that anybody is encouraged to do reviews and Simon or I will ack/push the patches.
14:10 walterbender_ silbe: +1
14:11 tch is anybody literally and singular? :)
14:11 silbe tch: maybe "everybody". I'm not quite sure about the subtle linguistic differences. ;)
14:12 tch so you just need one review from someone.
14:12 walterbender_ silbe: maybe you should say it in German, so it is at least clear between you and Simon :)
14:12 tch haha, no please..
14:14 silbe This is a short-time compromise to get us going again. In the long run we'll need to work out how to better collaborate during code development (i.e. even before the review stage as it is now).
14:14 tch silbe: +1
14:15 walterbender_ silbe: patch review is a different beast than architecture :)
14:15 silbe tch: to get a patch pushed to mainline, you'll need an Ack from Simon or me. If somebody else already did a thorough review, Simon and I don't necessarily need to do a detailed review as well, but can just Ack the patch.
14:17 walterbender_ mas facile
14:17 m_anish silbe, makes sense
14:17 silbe walterbender_: that's another topic we'll need to tackle soon. But it's not what I meant. We need to work out how to better work with "less experienced" developers. That's very important in the long run.
14:17 tch silbe: i get it :) sounds fine
14:18 silbe walterbender_: "mas facile"?
14:18 walterbender_ silbe: simple
14:18 silbe k
14:18 tch walterbender_: haha
14:18 bernie silbe: looks like a disk crash :-(
14:19 walterbender_ silbe: re working with less experienced developers, it would be good to hear from alsroot re his experiences
14:19 m_anish silbe, we have some programmers in india who are itching to contribute :)
14:19 silbe, maybe i'll discuss that later in the meeting
14:19 CanoeBerry has quit IRC
14:20 silbe and even within the "core team" we could benefit from tighter collaboration. But as I said that's a long-term task, so I'd say we discuss is outside the meeting.
14:20 tch silbe: i am also trying to gather some enthusiasts from many engineering colleges :)
14:20 m_anish: ^^
14:20 walterbender_ aside to alsroot: my journal problem was due to file permissions getting messed up in the restore... all is well in datastore land :)
14:20 silbe sounds awesome, m_anish and tch! Let's talk about it after the meeting.
14:21 m_anish tch, yeah :) maybe we can share experiences
14:21 silbe, sure
14:21 alsroot walterbender_: fine, anyway ds wasn't changed from 0.88
14:21 walterbender_ aside to tch: the UNA discussions are heating up :)
14:21 tch walterbender_: :D
14:22 silbe has anyone remaining questions re. patch review or should we move to the next topic?
14:22 tch silbe: i do,
14:23 silbe tch: fire away
14:24 tch silbe: (probably concerns alsroot too): after you are done with the patchwork, is it possible to invest some time on having a dextrose branch for 0.88.1 + dextrose patches ?
14:24 silbe:  for example, for the jhbuild env
14:26 silbe, alsroot: or any other env,
14:26 m_anish back in a moment
14:26 silbe tch: you should talk to bernie about that. He thinks it's a bad idea to have a git repo for the Dextrose sources and that we should use patch sets instead.
14:27 tch silbe: i see, ok i will talk to bernie about that. Is just that its been painfull lately its too "rambo" work lol..
14:27 silbe I'll use git for patch wrangling, but intend to only "publish" a patch set (exported from git, committed to the dextrose repo).
14:29 bernie tch: yes, I think the way it should be done is that you as a developer do not have to care. you work on master as you always did
14:30 m_anish is back (sorry)
14:30 bernie tch: then you send you patch to the buildmaster (or to silbe, now that he maintains the patch queue). and that person will massage the patch until it applies on top of the others
14:30 silbe Dextrose-2 patches -> smparrish, Dextrose-3+ patches -> me (via dextrose@)
14:30 dfarning bernie does any major project develop that way?
14:31 bernie tch: when I was doing it, I was just adding it to the spec file, and sometimes I had to fix the rejects manually until the patch applied cleanly.
14:31 dfarning: yes, fedora.
14:32 dfarning: we do not develop, we just do bugfixes on top of an upstream project.
14:32 dfarning: if we do not intend to ever rebase on upstream (i.e. a fork), then it's better to have a separate git repository
14:33 silbe bernie: so new features should go through upstream instead of getting added to Dextrose right away?
14:33 dfarning bernie, I'll read more about that. I disagree that dextrose does not develop.
14:34 bernie silbe: it doesn't matter whether it's a new feature or a bugfix. but we should not carry around large deltas. it's a big cost for us, regardless of which tools we use. because the tool doesn't actually matter that much.
14:36 dfarning: yes, it does develop. but this is more because upstream was dead(ish) and unresponsive to the deployment's needs. new development shouldn't be the focus of dextrose imho
14:37 silbe bernie: certainly. The question is a) whether we apply all patches to Dextrose first (and rebase if we need to modify them due to upstream requests) or channel some of them through upstream and b) when we throw out patches rejected upstream.
14:37 bernie silbe: whichever lets us have stable rpms faster and with less effort.
14:38 silbe: upstreaming things is an important, but secondary goal
14:38 silbe ok
14:39 bernie silbe: the was I was doing it is that I was submitting patches upstream for review first, but also apply them to dextrose for testing immediately without waiting.
14:40 silbe bernie: yep, that's what I intend to do as well (for patches I accepted).
14:40 bernie silbe: I think it's important that we maintain a process that lets us fix bugs quickly and not waste any time on juggling patches or complex interactions with upstream. we should stay user-focused.
14:41 dfarning let's switch to user feed back from anish.
14:41 silbe dfarning: +1
14:41 bernie yes
14:41 m_anish tch, pls change the topic :)
14:42 tch #topic Anish welcome to Paraguay and in-situ feedback.
14:43 m_anish This has been an eventful week for me, moving from India to py and the weather in py going from 15 degrees to 35 :)
14:43 tch in one day!
14:43 m_anish anyway, I've had useful discussion with roberto (rralcala, tch) and bernie on irc about the way things work here
14:44 I summarized them in a mail I sent on dextrose-ml last night
14:44 bernie did not read it yet
14:44 m_anish bernie, np, i'll just discuss it here anyway
14:45 so, on the tech front, the things they want most are (1) for dextrose to be stable (2) the yum updater to work
14:45 but i guess this is already known here...
14:46 also, it would be useful if we could figure out ways to reduce the time spent between deploying the s/w and discovering issues with it
14:46 tch 's notification system would be a step in that direction. there is probably room for much more.
14:47 silbe m_anish: how do you think the notification system would help?
14:47 and is there a way to get the Formadores to channel bug reports to us?
14:48 m_anish silbe, well, the formadores won't run to us (or run to us with more info) everytime something crashed. maybe automated bug reporting to go with this.
14:48 tch silbe: bernie and I tried that, more than once..
14:48 m_anish silbe, hmm, i'm still thinking of something on that front
14:48 silbe m_anish: Are actual crashes common in the field?
14:49 tch m_anish: +1 on automated tool
14:49 m_anish silbe, actually I don't know that very much. tch and bernie would know more
14:49 tch silbe: very
14:49 silbe: i mean, all kind of bugs.. mostly when you are at testing stage
14:49 silbe tch: Very interesting. XO-1 (=> low memory) or XO-1.5?
14:50 but automated reporting will only work for crashes
14:51 maybe Tracebacks after we cleaned up our code not to report tracebacks for potentially expected cases (e.g. OHM missing)
14:51 m_anish silbe, i'm not saying that 'automated reporting' is the best solution. I'm saying that there is scope for feature development in this area.
14:51 silbe and Tracebacks alone are rarely useful, we need to know how the problem was triggered.
14:52 m_anish: I certainly agree there is "scope". I just wonder whether it's worth a lot of effort.
14:53 CanoeBerry <CanoeBerry!~CanoeBerr@c-98-217-180-254.hsd1.ma.comcast.net> has joined #sugar-meeting
14:53 silbe but if anyone comes up with a reasonably simple patch, we should give it a try (for a limited number of "beta testers" to avoid getting overloaded by false positives).
14:53 bernie silbe: I think our best chance is to send out m_anish to the schools to collect bug reports directly from users
14:53 silbe: the formadores can report some kind of problems, through the technical team of pyedu
14:54 m_anish bernie, silbe one of the things rralcala told me that the formadores are so overburdened by their work, they can hardly give useful feedback
14:54 tch bernie: that did not work in the past bernie.. ;/
14:54 m_anish so making their life easier is a priority for us
14:54 walterbender_ tch: what did work in the past?
14:55 bernie tch: well, one would expect the formadores to have acquired more technical experience by now?
14:56 tch: and also, some formadores were actually good at reporting bugs. like ricardo, for example.
14:56 tch bernie: not sure if we should depend on that to happen, before starting to get real/useful feedback
14:56 silbe bernie: ok. maybe that works well enough that people at schools get encouraged to interact more closely with us. A single person at each school should suffice, and it doesn't have to be a Formador(es?). They just seemed like a natural choice as they're a) directly "on the ground" and b) a reasonably small set so they can interact directly with us. But if they're already overloaded...
14:57 bernie m_anish: yes, I think it would be best if you could visit the classrooms and collect problems directly from teachers and students
14:57 tch: sure, agreed. (see above)
14:57 m_anish silbe, in my experience, having a lot of info (even if all of it is not useful) is better than having none
14:58 bernie silbe: yes, this worked last year. not all schools will choose to participate, but some definitely did last year.
14:58 m_anish bernie, i'm still figuring that out as i get to know more but i've noted down your advice :)
14:58 bernie silbe: with so many teachers, it's not hard to find a few who are enthusiastic about testing new builds.
14:58 silbe m_anish: automated processes can easily drown you, making it hard to find the useful information. We have that problem already with systems-logs@ :)
14:58 walterbender_ bernie, tch: how about recruiting some students to take on some of this burden, e.g, Benedicto and his peers
14:59 m_anish bernie, are you talking about the schools you went to?
14:59 bernie m_anish: the two schools that were most helpful last year were Daniel Ortellado and Teniente Aquino. Walter visited both of them.
14:59 walterbender_ bernie, tch there must be many students who could do this at this point
14:59 m_anish bernie, i guess you know, but there would be 27 more schools on top of the 10 we have when the new laptops are deployed
14:59 bernie, ok!
14:59 tch personally (and i hope someone probes me i am wrong ASAP) I think that it will not work so well.. every time we land that kind of responsibility on someone (outside of the main tech  team) our feedback chain breaks.. is sad but is true..
15:00 bernie walterbender_: yes, that's what I did. give new versions to geeky students who would report bugs to me.
15:00 walterbender_: but this required me to go often to caacupe... because communication over email and irc does not work very well with teachers and students.
15:00 walterbender_ bernie: I don't think there is any way around that...
15:00 tch i think we should aim for that.. but not depend on that.
15:01 walterbender_ tch: I think you need redundant mechanisms
15:01 bernie m_anish: yes, but those new schools will have very different problems: power, networking, schoolserver, hardware maintenance...
15:01 m_anish: they won't be very useful to debug dextrose for at least one year
15:01 walterbender_ in .uy, they have Ceibal RAP which visits schools to complement the more formal processes of Ceibal
15:01 bernie m_anish: I strongly recommend working with experienced users who can tell us about regressions (it used to do this, now it doesn't...)
15:02 silbe bernie: what's the reason communication via email/IRC doesn't work well? Bad/no internet connection, different culture? Is there a way we could improve it from our side (snail mail, VoIP, different style of interaction, whatever)?
15:02 m_anish walterbender, agree we could probably try both (automated and going to school, collecting info from willing teachers/students) and see if they work well in tandem
15:02 tch m_anish: +1
15:02 walterbender_: +1
15:02 walterbender_ what about a Saturday club like the Tortegart, Scratch, and Etoys clubs,  only dedicated to bug tracking?
15:03 bernie silbe: I did anything I could to solve the technical issues... I included irc in the builds, made it join #olpc-paraguay automatically....
15:03 silbe: but at the end of the day the problem is that users are not used to chat on the internet.
15:03 silbe: of 15 formadores, only 2 did show up sometimes. and they would not stay online long enough to get an answer from us most of the times :-)
15:03 tch bernie: exactly.
15:04 walterbender_ I bet you could find a great team of kids who'd be interested, as well as a formadora or two...
15:05 they have a growing mechanism for these informal-time clubs -- classroom time is too busy to use for debugging
15:05 bernie silbe: children were similarly unable to communicate online beyond "hello, how are you?". the only way to collect feedback from users is taking the bus, trust me :)
15:05 walterbender_ tch: I'll bring it up with Pacita
15:05 tch walterbender_: :)
15:05 bernie walterbender_: yes, I agree. it's important to work with the users who are enthusiastic about testing new software.
15:06 walterbender_ m_anish: if you ran a Saturday club in Caacupe, (1) you'd get great feedback and (2) you'd have a great time
15:06 m_anish i think it would be worth having an activity that allows to easily report problems, maybe include screenshots
15:06 silbe bernie: taking the bus doesn't scale for us, that's why I'm trying to figure out other ways. Either directly (via internet) or indirectly (e.g. finding university students willing to visit a school once a week).
15:06 m_anish walterbender_, ok :)
15:06 dfarning walterbender please don't bypass the lines of communication that are in place m_anish can present the idea to scs and roberto
15:06 bernie m_anish: I don't think that was the actual issue
15:07 walterbender_ silbe: in the next few months, we'll have many more university students on board...
15:07 silbe walterbender_: great!
15:07 bernie m_anish: users were able to understand how to use trac (it's an easier webapp than gmail, after all). they just would not describe the problem properly
15:07 m_anish bernie, hmm
15:07 tch bernie: and after a while they stopped doing it
15:08 bernie: except for people for mentioned
15:08 bernie m_anish: they could not distinguish software from hardware, you see? they'd say things such as "I can't see the battery any more" when the frame key on the keyboard is stuck.
15:08 walterbender_ bernie: at one point in .uy, they were going to work on a kid/teacher-friendly trac... don't know if it ever happened.
15:08 silbe BTW (sorry, getting off topic), has anybody followed up with the government guy who wanted to fund SoaS work? Depending on the details, Dextrose might work well enough for them.
15:09 bernie tch: yes, indeed
15:09 walterbender_ silbe: which government?
15:09 silbe walterbender_: US I think
15:09 bernie silbe: uh?
15:09 that would be great
15:09 silbe goes looking for the mail
15:10 m_anish bernie, tch hmm, so what do you think would work in addition going to school
15:10 bernie so, guys. what *will* work is: m_anish goes to caacupe every once in a while to talk with volunteer testers and give then new releases
15:10 m_anish bernie, we could have yum-testing repos when we have the yum-updater up
15:11 bernie SMParrish_away: I think we should drop the concept of weekly or bi-weekly releases. we should have releases when there are bugfixes we want to test, as often as needed for testing.
15:11 that is, be test-driven.
15:11 m_anish: sure
15:11 walterbender_ bernie: +1
15:11 m_anish to beta test the 'new & shiny' and get great feedback. but that is a differernt topic
15:11 tch bernie: i agree on that, ALWAYS they come thought the updater..
15:11 bernie m_anish: we already have the right structure for this (release and testing dirs in the dextrose download site). we're just missing a cronjob
15:11 silbe Message-ID <AANLkTik-fqTD6U83FdOAi=SpsBd+1pkYxdXu_WhNP4MW@mail.gmail.com>
15:12 tch bernie: the method of going there and re flashing machine for each bug fix is to intrusive
15:12 bernie tch: yes, this will remove the need to go there for "giving" them the new software.
15:12 m_anish bernie, ok, re: that we are planning to start testing the yum updater starting this monday
15:12 with the ~10 xo's we have at the pyeduca office
15:12 silbe http://lists.sugarlabs.org/arc[…]ember/002047.html
15:12 bernie tch: the downside is that it will be hard to tell which version of sugar the users were running. we should remember to bump the release number on each build.
15:13 m_anish: great
15:13 tch bernie: new fixes, new rpms, new release number, makes sense..
15:14 bernie silbe: the email sounds a bit odd... did anyone followup?
15:14 m_anish can we continue this on m-l and switch topics?
15:14 tch bernie: and if the updater works as intended, we won't have to worry, because they should always get the updates if they are conencted to the XS
15:14 bernie walterbender_: maybe you could answer this government guy?
15:14 m_anish: +1
15:14 tch can i change topic then?
15:15 walterbender_ silbe: I was a bit suspicious of that email
15:15 m_anish tch its the same main topic
15:15 i'm switiching to non tech aspect
15:15 silbe ok, let's continue after the meeting or via email
15:15 tch m_anish: ok, go ahead :)
15:16 m_anish one of the things we're trying to think of (and me and arjs talked about this extensively) is how to get the edu. team more involved in mainstream SL pedagogical discussions...
15:17 the problem with that is bridging the communication gap (probably)
15:17 they would have a pretty good way of analyzing and figuring out what works in an 'education & learning' sense IMHO
15:18 secondly, how can we get similar teams to collaborate across deployments (py, uy ...) but i have yet to talk with the edu team re: this
15:19 tch probably emulating what we did in the technical side,
15:19 m_anish that's pretty much what i wanted to say
15:21 i'm hoping everyone will have plenty of thoughts on this :)
15:22 dfarning m_anish, please feel free to work on the educational parts of the puzzle with groups like SL.  For the next 4 months, as AC, we are going to focus on the technical improvements.  After we have established momentum on solving technical problems we will start looking at other issues like educational efficacy.
15:23 bernie is still distracted by savannah issues
15:23 m_anish dfarning, ok
15:23 dfarning m_anish, our resources are rather limited..... and we have to bootstrap the business.
15:24 m_anish tch, i'm done with what i had to say. I would like to have discussion re new developers but not sure whether it should be part of the meeting.
15:25 tch i have one more topic
15:25 then we can discuss about it ? :)
15:25 m_anish dfarning, i'm having this discussion because we might end up having a tech soln. to this..., but I get what you're saying.
15:25 tch, sure go ahead, you're the meeting boss :)
15:26 tch # feedback and bug reports from ceibal
15:26 ops!
15:26 #topic feedback and bug reports from ceibal
15:27 ceibal folks seems the doing good tests lately, and they keep reporting the problems that affects also to py or any other dextrose user
15:27 bernie dfarning: when were you planning to come to boston?
15:28 tch and as i mentioned bernie, i am a little worried that we haven't move much in that direction lately,
15:29 bernie dfarning: regarding new developers, I think we have enough people on the team if we agree to make all of them focus on stability, with a tight loop between anish testing and tch + silbe fixing bugs quickly.
15:29 oh, and alsroot
15:30 tch: right. they don't actually seem to be reporting *all* the problems they see. only those they don't know how to solve :-/
15:30 tch: we should probably ask them more directly to send us their bugfixes and share their bug list (I'm sure they have one internally)
15:32 tch: they also seem to have their own rpm repository and their own release schedule... at this time, the uruguayan build is more like a fork of dextrose 1 than something we're doing together.
15:32 tch bernie: wasn't that the idea?
15:32 bernie tch: we should have someone (you?) visit them for a few days to resync
15:33 tch bernie: i am going to uruguay this week, and ill visit them :)
15:33 bernie tch: as with the end-users, communication over the internet is insufficient. only face-to-face meetings are effective.
15:33 tch: GURAIT-TO!
15:34 m_anish: unrelated, but important: it would be good for you to get a good spanish teacher. the colleagues in the office will talk to you mostly in english, you won't learn much from them.
15:34 m_anish bernie, yes, i'm planning to start on that very soon
15:34 tch bernie: i just don't want to get into bad assumptions, i think they have increase their community interaction a alot
15:35 bernie m_anish: good. I hope pyedu can pay expenses.
15:35 walterbender_ tch: lean on Emiliano. He is a good guy and very pragmatic
15:35 silbe can we convince Ceibal & co to sponsor travelling for regular in-person meetings with them?
15:36 tch walterbender_: ok :)
15:36 bernie tch: yes, me too. I don't think they *want* to work in isolation.. it just happened because both you and me stopped being in contact with them all the time.
15:36 silbe: probably yes
15:37 silbe: they did sponsor me 6 months ago
15:37 silbe: for one week, because I had to leave for mozambique
15:37 tch bernie: i try to keep in contact :), for example now i switched to bug fixing mode again hehe
15:39 i just think the interaction should be bi-directional if they report bugs and no one effectively helps then, then we are failing..
15:40 bernie tch: yes indeed
15:40 silbe bernie: great! we should talk about that in January or February (next few weeks are going to be busy with "family" stuff AKA christmas/new year's eve).
15:41 bernie tch: I think what worked is that daniel ortellado was testing for real, and we were fixing their bugs for real.
15:41 tch: if one of the two fails, the feedback loop breaks.
15:41 silbe tch: right. We have that problem in general on the upstream side. :-/
15:41 tch if there is anything you guys want me to also do there, let me know
15:41 gangil has left #sugar-meeting
15:42 bernie silbe: in my case, they had to pay only a cheap ticket from paraguay... not sure they'd pay a ticket from germany :)
15:42 m_anish :)
15:42 silbe I'd say we continue that discussion after the meeting.
15:42 tch kk, is there any other topic concerning this meeting or should i end it?
15:42 bernie silbe: I think it would make more sense for you to visit paraguay, if possible. because uruguay is too large and too well structured to be representative of a typical olpc deployment.
15:42 silbe bernie: I wasn't talking about going there myself. Just that I'll need to focus on other things within the next few weeks.
15:43 tch silbe: you should come, seriously... all the core members should visit deployments..
15:43 silbe bernie: Bine prefers Peru ;)
15:43 bernie silbe: you'd only see an office and work with regular programmers like you, most of the time. in paraguay, instead, you'd be more in contact with schools, support engineers, educators, etc.
15:43 silbe: don't know about peru, but it might be the same of uruguay, just more disorganized.
15:44 tch any other topic? counting down..
15:44 5..
15:44 silbe I'd love to go to some deployment after I finished my studies and moved to a new place (we're currently living in dormitories and have to leave when I finish my studies, i.e. at the end of the semester).
15:45 bernie silbe: well, if you had time, I would really like you to spend a few weeks at a deployment to get a sense of how users interact with sugar and how they think.
15:45 silbe 4
15:45 tch 2... (obviously not seconds ;))
15:45 #endmeeting
15:45 meeting Meeting ended Sat Dec 18 15:45:49 2010 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot. (v 0.1.4)
15:45 Minutes: http://meeting.sugarlabs.org/s[…]-18T14:04:53.html
15:45 Log:     http://meeting.sugarlabs.org/s[…]10-12-18T14:04:53

Minutes | Index | Today     Channels | Search | Join

Powered by ilbot/Modified.