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

#olpc-admin, 2009-01-06

Index | Today     Channels | Search | Join

All times shown according to UTC.

Time Nick Message
16:02 hhardy #topic Agenda
16:02    *  Ideatorrent is set up (Dogi)
16:02    * Improving documentation process and availability wiki git
16:02    * Need cluebot or similar for wiki spam issue
16:02    * Sugar and library list admin help needed
16:02    * backups
16:02    * New business
16:02 #topic Ideatorrent is set up (Dogi)
16:02 dogi hi all
16:02 hhardy Dogi has set up ideatorrent per ticket by Seth/is4
16:03 dogi http:/idea.mogidi.net
16:03 hhardy Do you want to say something about how you envision us using this for OLPC?
16:03 dogi please do all register an account
16:03 http://idea.mogidi.net/drupal5/user/register
16:05 so in the meantime I will explain the purpose of this service
16:05 at first there are developer and there community
16:06 and this is a tool build for the community to give feedback to developer ...
16:06 and sysadmins ... LO
16:07 cjl dogi have you asked any developers whether they wanted this?  They seem to prefer Trac.
16:07 dogi no it is build for all non devel
16:08 cjl It might be nice for other sorts of ideas, but I hate to see tools proliferate to the point they overwhelm those who are supposedly using them.  
16:08 dogi it has rss for all developer ... which i think its nice
16:09 cjl Since getting involved wit hOLPC, I've had to master IRC, wiki, RT, and I'm learning Trac, and I don't even write code. . .
16:09 I find it fun to learn new tools, others maybe less so.
16:10 hhardy So as of now, when a user has a suggestion they might post it to the devel list or enter a ticket for an enhancement into trac
16:10 How would that process work differently with this tool?
16:10 cjl or if not code related writes it up on wiki
16:11 dogi cjl please create bevor u talk an account
16:11 so that i can give u the needed right to show u ...
16:11 cjl *sigh*
16:11 goes t ocreate yet another account. . .
16:11 dogi sorry
16:12 :)
16:13 cjl ok, i'M IN
16:13 dogi thx
16:13 cjl dogi: /me switches t oinside voice.
16:13 dogi then create ur idea proposual ,,,
16:14 cjl um, need to change "Change my Ubuntu QA email or password "
16:14 dogi or go to sandbox to promote a waiting idea ...
16:15 yes i know that idea is not finished
16:15 thx
16:15 i gave u the right of moderator there
16:15 so u can delete ... :)
16:17 http://kuku.laptop.org/winpng/
16:17 hhardy I suggest people look this over and comment to the olpc-sysadmin mailing list?
16:17 dogi this are picture i did
16:18 hhardy anything further on Ideatorrent?
16:18 dogi http://wiki.laptop.org/go/User:Dogi/Ideatorrent
16:18 hhardy #topic Improving documentation process and availability wiki git
16:19 There's not much new to report on this but we've seen a few instances of wiki downtime where having the docs on the wiki which is down is mighty inconvenient
16:19 is there anyone who is more than a git novice?
16:19 here now I mean?
16:19 adric still likes Subversion :)
16:20 hhardy I will try to draft Bernie if nobody else volunteers
16:20 dogi is novice ... :P
16:20 cjl cjl votes for _bernie
16:20 cjb hhardy: hm, how are they related?  (the wiki isn't in git.)
16:21 cjl cjb push some docs t oa git repo, avail when wiki is down.
16:21 adric As an alternate for sysad docs, git has been suggested by several parties
16:21 hhardy the idea is make a static copy of the wiki using something like wget then check that copy ito git
16:21 hopefully via a cron job
16:21 Dogi came over and says mstone has some ideas on how to do this
16:21 adric prob want to dump the wikitext rather than wget the html, but yeah
16:22 mstone has some good ideas on a /root/docs folder on all managed servers
16:22 cjb how to create a git repo and put a file in it?  all of the git tutorials cover that; I could point you at one.
16:22 m_stone Do I hear my name being taken in vain?
16:23 hhardy yes
16:23 not in vain however
16:23 since you responded
16:23 dogi hi
16:23 cjl cjb more on how to automate regular push
16:23 dogi m_stone :)
16:24 adric m_stone the stuff we chatted about a couple weeks back for server docs .. local store and git clones
16:24 cjb ah.  while true; do sleep 3600; wget laptop.org/wiki/path -o wikifile; git commit -a -m 'latest version'; git push; done
16:24 m_stone cjb: enjoy the html you get from that.
16:25 hhardy #IDEA <@cjb> ah.  while true; do sleep 3600; wget laptop.org/wiki/path -o wikifile; git commit -a -m 'latest version'; git push; done
16:25 who is interested to help with this?
16:26 m_stone hhardy: the issue isn't really who wants to help with it. the issue is whether we can agree on a system that means basic requirements.
16:26 *meets
16:27 the internalwiki is unhelpful because it's inaccessible. the teamwiki is unhelpful because there's no consensus on using it. the public wiki is unhelpful because there's a tiny bit of private data.
16:27 etc.
16:28 however, I think that everyone agrees that, in order to admin a system, one needs to have an account on it.
16:28 cjb ah.  while true; do sleep 3600; wget laptop.org/wiki/Special:Export/path -o wikifile; git commit -a -m 'latest version'; git push; done
16:28 ^ returns wikitext.
16:28 adric cjb; +
16:28 hhardy #IDEA  <@cjb> ah.  while true; do sleep 3600; wget laptop.org/wiki/Special:Export/path -o wikifile; git commit -a -m 'latest version'; git push; done
16:29 m_stone consequently, I think that having an account on a machine should be sufficient to give you read + comment access to all public docs about the machine and that having root on the machine should give you the rest.
16:29 cjb this is supposed to work too:  http://wiki.laptop.org/go/inde[…]vities&action=raw
16:29 (but doesn't seem to)
16:29 m_stone I think the laptop.org/wiki part is what's wrong.
16:30 adric "Forbidden Raw pages must be accessed through the primary script entry point."
16:30 m_stone cjb: I don't actually care whether the docs are version-controlled or not because I can handle that independent of anyone else with an equivalent for-loop.
16:31 adric can we make the access control work the way described above? Easily?
16:31 m_stone e.g. I can put a cronjob on my laptop that does for s in $SERVERS; do rsync -avz $s/root/docs $server-`date`; done
16:31 hhardy we dont need all pages only the VIG and sysadmin docs
16:32 m_stone or equivalent git magic if I happen to like git.
16:32 cjb m_stone: I agree
16:32 m_stone as for adric's question:
16:32 here's a rough draft.
16:33 cjb The problem isn't at all that someone needs to show people how to use git, it's that you need to know how to get the right data out of mediawiki in an automated way including for our wikis with access control
16:33 m_stone cjb: consequently, I think that we should store the data in a filesystem (which we all know how to use) and push it into a wiki if that's really needed.
16:33 (I happen to think it's not)
16:34 cjb hm, let me try to understand that
16:34 adric Well, there is the wiki FS fuse module ...
16:34 cjb so you want Machine:* to disappear?
16:34 that was pretty convenient
16:35 m_stone cjb: I tend to think that    ssh $FOO cat /root/docs/purpose is fairly convenient too.
16:35 cjb What if I can't ssh to foo?
16:35 m_stone two answers:
16:35 a) then you shouldn't be reading its docs.
16:35 cjb What the way I enter foo is by sshing to solar and running "vserver enter"
16:35 ?
16:36 What if foo lives in a safe?  :)
16:36 in other ways, I think that (a) is a cop-out.
16:36 m_stone cjb: then you go to the safe in order to read its docs, or you /republish/ those docs.
16:36 cjb s/ways/words/
16:36 m_stone cjb: I did say that there were two answers.
16:36 cjb m_stone: on some kind of.. wiki, or something?
16:36 m_stone you did sort of cut me off.
16:36 cjb it's true.  I'm sorry.
16:37 m_stone cjb: I don't care on what since it's not authoritative.
16:37 the authoritative docs should live on the machine.
16:37 adric ... and be cloned elsewhere
16:37 m_stone anything else seems to me to be a central point of failure.
16:37 hhardy m_stone: how does that help us when machine X is down?
16:37 ah IC
16:37 m_stone hhardy: because you have non-authoritative backups, don't you?
16:37 since they're so easy to create now?
16:38 adric And it's the clone that I wonder about the access control of, is all.
16:38 cjb hhardy: it doesn't, so we republish that information in a wiki, and then everyone just uses the wiki anyway because it's more convenient
16:38 m_stone cjb: that could be, but I don't think it will be.
16:38 cjb: it would obviously indicate that my ideas are wrong if it happened.
16:39 i.e. that centralization is fine and good and that we worked out some way to publish the public bits.
16:39 which I don't know how to do sanely with our wikis.
16:39 but which I certainly know how to do with filesystems.
16:39 cjb I'd like your ideas a lot more if I didn't usually look at Machine:* pages when a machine is down or inaccessible to me.
16:39 m_stone cjb: I'd like your ideas a lot more if I had some confidence that the Machine:* pages were going to be available when I need them.
16:40 and that we knew how to make the public parts public.
16:40 cjb I see
16:40 m_stone I think the first problem is solvable.
16:40 it's ugly because mediawiki is ugly, but it's definitely solvable.
16:40 just with some annoying scripting.
16:40 I don't know how to solve the second problem.
16:40 (with your system).
16:40 I know how to solve it with a filesystem.
16:41 e.g. by making a separate "public" folder to clearly demarcate what is public and what is private.
16:41 cjb == Public == ?  :)
16:41 m_stone cjb: how do you implement that so that members of the public can see it?
16:41 how do the logins work?
16:41 cjb how do you implement yours?
16:42 it's on a filesystem on production machines with restricted accounts.
16:42 I don't see the difference between the two yet.
16:42 m_stone I'd copy them off.
16:42 cjb then you can copy them off the wiki too!
16:42 hhardy I'm calling timeout to this topic in 4 min
16:42 m_stone either via my account directly or via a restricted-login public account with private passwords.
16:42 cjb: yes, I said that. I just said that it's a lot harder.
16:43 as in, I don't /know/ a solution off the top of my head.
16:43 cjb okay.  I remain unconvinced, but that's okay.
16:43 this is a good conversation to have recorded, at least
16:43 m_stone you'd rather that trust that I can write a parser which correctly extracts the public bits of your wiki pages?
16:43 hhardy I suggest we continue the discussion at http://wiki.laptop.org/go/OLPC[…]rastructure_Group and/or on the olpc-sysadmin list
16:44 cjb laptop.org/teamwiki/go/Machine:Solar/Public?  :)
16:44 hhardy I suggest that as a first approximation, we do *something* so that docs are still available when wiuki is not
16:44 cjb hhardy: copy them into a text file.
16:44 m_stone cjb: just seems like a big waste of time to me. (though I know that my proposal seems the same to you)
16:44 adric Wait, are allowed to act? I thought this was a discussion -only group. :D
16:44 hhardy *bonk* adric
16:44 cjb m_stone: the proposals seem equivalent to me, yeah.
16:45 hhardy and so that VIG has access to machine: pages at least most of them
16:45 cjb (in the regard of having demarcated areas for public and private info)
16:45 m_stone cjb: the wiki is just so hard to script.
16:45 (comparitively)
16:45 cjb yeah
16:46 m_stone and I hate that I have to log in to it every 10 minutes.
16:46 I also despise using firefox's text editor.
16:46 cjb hhardy: honestly, I'd spend a half hour recording all the information the VIG might want currently in a single e-mail and sending it to the list
16:46 m_stone all of these things are fixable, /maybe/, but it's /work/ that I don't feel motivated to do.
16:46 hhardy time
16:46 cjb: certainly better than nothing
16:46 cjb m_stone: likewise
16:47 hhardy #IDEA <@cjb> hhardy: honestly, I'd spend a half hour recording all the information the VIG might want currently in a single e-mail and sending it to the list
16:47 m_stone cjb: whereas I'm confident that I could manage to put all the data onto the write machines /once/ in just a few minutes.
16:47 *right
16:47 hhardy #TOPIC Need cluebot or similar for wiki spam issue
16:47 dogi :)
16:47 hhardy brief discussion on this
16:47 we are having more than usual amount of defacement and blankings on wiki
16:47 cjl There have been spasms of fairly intense wiki vandalism
16:48 hhardy there's general agreement I think that cluebot is a good measure
16:48 cjl then they subside, and eventually recur.
16:48 hhardy #LINK http://en.wikipedia.org/wiki/User:Cobi
16:48 cjl the "community" patrol aspects of the wiki don't work as well as you might like.  It tends to be one or two users doing most of the work.
16:49 _sj_ aww
16:49 cjl I did some analysis of this and posted it on the main wiki a while back onder a page called 'Bot Interest Group"
16:49 _sj_ that's a factor of the [# of editors / popularity of wiki] ratio
16:49 we need to work on that.
16:49 give newbies regular wiki tasks as one of the s-g and vig and other efforts
16:49 RC patrol is a great simply newbie task.
16:50 dogi http://code.google.com/p/falebots/
16:50 hhardy _sj_: do you prefer this approach to cluebot? or in addition to?
16:50 cjl _sj_: It is also desirable sometimes to investigate (and perhaps even implement) suitable technical solutions.
16:50 hhardy #LINK <@dogi> http://code.google.com/p/falebots/
16:51 _sj_ hhardy, you never fail to surprise me.
16:51 we need cluebot.
16:51 no large wiki should be without it.
16:52 [rc is 'recent changes']
16:52 we have 25k pages on the wiki, which is more than middling size
16:52 cjl ffm had initially made some claims of knowing Cobi via Cluenet, but recently says he hasn't been active on that network for a while.
16:52 _sj_ cjl, absolutely
16:53 I have been the longest proponent of getting cluebot to work
16:53 this is my top infrastructure priority as far as the wiki goes
16:53 I'm sorry that I don't know how to do it myself.
16:53 cjl An ideal situation would be to have Cobi act in a consulting vol (if someone jknows and can recruit him).
16:53 hhardy and he's been advocating it for a long time too! :)
16:54 _sj_ and it seemed as if someone with more scripting experience was interested in the problem
16:54 we can also write to the cluebot fans and ask for someone to lend a hand
16:54 not just cobi
16:54 hhardy #LINK http://rt.laptop.org/Ticket/Display.html?id=1871
16:54 _sj_ any of the people working on the recent rewrite, say
16:54 hhardy its by far the oldest active sysadmin ticket I believe
16:54 cjl My only concern about Cluebot as it exists currently is it's potential lang-en bias
16:55 hhardy BTW is there anyone who does not have access to the sysadmin queue on rt who wants/needs it from VIG?
16:55 cjl there is Crispy-something working on hte ANN rewrite (apparently).
16:56 hhardy I suggest updating that ticket if you know something useful, and take general discussion to the list
16:56 cjl I have a cvague interst in ANN's and have posted ideas on Cripsy's page as to how to gather known good-bad sets of edits for the training sets.
16:57 but certainly not the coding skills.
16:58 hhardy SJ has had this opportunity posted since August and as its on the VIG page thought I'd mention:
16:59 #TOPIC Sugar and library list admin help needed
16:59 #LINK http://lists.laptop.org/piperm[…]ugust/000703.html
17:00 The sugar list has been moved to sugarlabs
17:01 SJ says library list could use another moderator
17:04 anything further?
17:04 _sj_ I need to set up some web forms and run scripts on their inputs
17:04 I've posted a couple of times about this to sysadmin@ but perhaps not to vig
17:07 hhardy There's an rt ticket for this but cant find atm
17:07 http://rt.laptop.org/Ticket/Display.html?id=23365
17:08 ok thats it for today thanks!
17:08 #LINK http://rt.laptop.org/Ticket/Display.html?id=23365
17:08 #endmeeting

Index | Today     Channels | Search | Join

Powered by ilbot/Modified.