OLPC:Volunteer Infrastructure Group/2008-10-07

Agenda

17:22 < hhardy> Need secretary
17:22 < hhardy> Wiki and website scalability/mirroring -- deployments + G1G1 are coming!
17:22 < hhardy> download.l.o scalability/mirroring Google, archive.org others?
17:22 < hhardy> Lfaraone committee on single-signon options
17:22 < hhardy> SAGE Sysadmin Ethics Statements
17:22 < hhardy> GPG
17:22 < hhardy> State of rt rt instances for Adric and private
17:22 < hhardy> Big Sister oncall schedule Big Sister tuning Big Sister maintainer
17:22 < hhardy> new business

Need secretray

17:16 < lfaraone> ---- Start of Meeting ----
17:16 < hhardy> oops I am not ops for some reason...
17:17 < lfaraone> ACTION ITEM: Give hhardy a raise
17:17 < lfaraone> (srsly, real meeting starts below)
17:17 < adric> lfaraone: Also, a dig at Linus for not making 2.7. And using git. there I'm done,
17:18 < lfaraone> TOPIC: Need secretary
17:18 -!- kimquirk [~kimquirk@can-olpc-nat.media.mit.edu] has joined #olpc-admin
17:19 < lfaraone> Anybody want that job? (secretary for those who just joined)
17:19 < dogi> what is the secretaries job involve?
17:19 < dogi> ups does
17:20 < lfaraone> dogi: Taking meeting minutes, which hhardy currently does AFAICT.
17:20 < hhardy> hmmm chanserv doesnt love me anymore
17:20 < lfaraone> dogi: Managing the calendar.
17:21 < lfaraone> hhardy: It dumped you for MemoServ. Long-distance relationships just don't work out :)
17:21 -!- gregdek [~gdk@nat-pool-rdu.redhat.com] has quit [Ping timeout: 480 seconds]
17:21 < lfaraone> dogi: You up for it?
17:21 < dogi> :)
17:22 < hhardy> bah oh well I will fix it later, I should be auto-opped on this chan
17:22 < dogi> i think cause i m not native english speaker i m not that good for that ..
17:22 < hhardy> heres the agenda
17:22 < hhardy> ---START---
17:22 < adric> dogi: you seem to do better than the Americans on help@ ;)
17:22 < hhardy> Agenda
17:22 < hhardy> Need secretary
17:22 < hhardy> Wiki and website scalability/mirroring -- deployments + G1G1 are coming!
17:22 < hhardy> download.l.o scalability/mirroring Google, archive.org others?
17:22 < hhardy> Lfaraone committee on single-signon options
17:22 < hhardy> SAGE Sysadmin Ethics Statements
17:22 < hhardy> GPG
17:22 < hhardy> State of rt rt instances for Adric and private
17:22 < hhardy> Big Sister oncall schedule Big Sister tuning Big Sister maintainer
17:22 < hhardy> new business
17:24 < hhardy> being secretary means at a minimum taking the meeting transcript, pulling out the action items, and making pretty using basic wiki markup like Seth, Mel and I have done with past minutes
17:26 < dogi> ok i ll try with the minutes this time ... ok?
17:26 < hhardy> sure
17:26 < hhardy> you can do last week too if yoyu like?
17:27 < hhardy> ACTION ITEM: Dogi to try preparing this week's minutes

wiki and website scalability

17:27 < hhardy> TOPIC: wiki and website scalability
17:27 < dogi> hhardy: i dont have the logs of that
17:27 < hhardy> last week Monday and Tuesday the wiki had 22,000 unique visitors from Uruguay
17:28 < hhardy> I will email the whole logfile if thats ok?
17:28 < kimquirk> hhardy: how do we all get the graphs of these visitors?
17:28 < dogi> ok
17:28 < hhardy> dogi I dont have your email
17:29 < dogi> stefan@untehauser.name
17:29 < dogi> ups
17:29 < hhardy> that was from google analytics from isforinsects
17:29 < hhardy> .name?
17:29 < kimquirk> isforinsects: can you post the link?
17:29 < dogi> stefan@unterhauser.name
17:29 < dogi> stefan@unterhauser.name
17:30 < lfaraone> hhardy: Well, has any work been done on Squid or other cache servers?
17:30 < hhardy> isforinsects: ping
17:30 < hhardy> yes squid reverse proxy was set up on weka
17:30 < isforinsects> woah...
17:30 < hhardy> the load average went down from 177 to 2 then
17:30 < dogi> hui
17:30 < adric> hhardy: unterhauser.name MX is 66.92.78.228
17:30 < isforinsects> kimquirk, analytics.google.com if you give me a google account I can give you access.
17:31 < kimquirk> kim.quirk@gmail.com
17:31 < kimquirk> (is that what you need?0
17:31 < lfaraone> isforinsects: (!meeting) btw tht isn't working for me either
17:31 < isforinsects> Most of the week after the spike it appears that we've lost all of UY's hits
17:31 < adric> Lost or absorbed?
17:31 < isforinsects> ffm246?
17:31 < lfaraone> isforinsects: ( I am able to go to the account, but I see nothing there)
17:32 < hhardy> dogi: sent
17:32 < isforinsects> lfaraone, try now
17:32 < isforinsects> adric, ?
17:32 < hhardy> they are going to the proxy now?
17:32 < cjl> works for me
17:32 < lfaraone> isforinsects: No Profiles found.
17:33 < lfaraone> isforinsects: try removing and readding ffm246 at gmail dot com 
17:33 < lfaraone> isforinsects: got it
17:33  * adric shrugs, adricnet@gmail
17:34 < lfaraone> Woah, half of our hits were from uruguy
17:34 < lfaraone> uruguay*
17:34 < hhardy> yeah it was a big surge, the previous week they had a week off
17:35 < hhardy> when they returned the went on a rampage for more activities
17:35 < hhardy> so...
17:35 < cjl> yes, but very low pages/visit count.  They click the links on Browse and then drop (usually)
17:35 < hhardy> we need to think about a scalable distributed infrastructure for upcoming deployments and particularly G1G1
17:36 < isforinsects> adric, added
17:37 < cjl> Why not ask people who really think deeply about such stuff (e.g. google)
17:37 < hhardy> cscott has investigated how the main english wiki is set up
17:37 < hhardy> we will ask him to talk to us next week, he is at the gnome thing according to richard
17:38 < adric> Oh, that's pretty data .. mmm
17:38 < hhardy> "some gnome conference"
17:38 < hhardy> since google is on our board and they have fat pipes with lot of dark fiber that makes sense
17:39 < hhardy> php is the main slowdown factor cscott says wikipedia has 140 dedicated pvp servers
17:39 < adric> CDN would be helpful for static content / cacheable pages
17:39 < lfaraone> hhardy: we can use coral's free cdn service...
17:40 < hhardy> ACTION ITEM: investigate CDN networks, such as coral

archive.org

17:40 < hhardy> we also shall look into archive.org
17:40 < hhardy> ACTION ITEM: ask SJ about setting up a VIG archive.org acct
17:40 < cjl> What sort fo traffic do big distros handle?
17:41 < hhardy> see http://en.wikipedia.org/wiki/Special:Statistics
17:41 < cjl> If it's all PHP, the mediawiki folks may have the only answers tha are directly relevant, but there are others that have some larger infrastructure (more users) than OLPC.
17:42 < hhardy> english wikipedia is in the top 20 alexa and neilson sites worldwide
17:42 < hhardy> so we arent probably going to exceed them
17:42 < adric> Anyway to give more visitors static (cached) pages is going to be a big win, since PHP/database is so much slower
17:42 < hhardy> lfaraone had a proposal for setting up and seeding torrents
17:43 < cjl> wouldn't think so, just suggesting tha others besdies mediawiki may have other sorts of solutions (for growing up oor).
17:43 < hhardy> which is a more efficient delivery method that static websites
17:43 < hhardy> *than
17:43 < adric> delivery of .. what would we torrent out?
17:43 < lfaraone> adric: OS images.
17:43 < hhardy> activities potentially
17:43 < dogi> hhardy: arrived
17:43 < hhardy> cool
17:43 < adric> Oh yes, please, for the OS images.
17:43 < lfaraone> hhardy: only big ones...
17:43 < lfaraone> hhardy: like the wikipedia one.
17:44 < adric> Activities? Yeah, maybe content packs and other big files
17:44 < hhardy> such as wikipedia activity
17:44 < adric> agree.
17:44 < lfaraone> hhardy: oh, IDEA: get a torrent client built into browse...
17:44 < hhardy> lfaraone: good
17:44 < lfaraone> I can propose taht for 9.1, and then it will be a seamless download...
17:44 < cjl> vast majority of the UY hits are Activities page
17:44 < adric> so thousands of the UY machines have daily internet access?
17:44 < cjl> 237K of 263K
17:45 < adric> If so that's awesome.
17:45 < lfaraone> hhardy: basically it's not that hard to set up a torrent server or even your own tracker.
17:45 < lfaraone> hhardy: although you will have to be seeding.
17:45 < adric> would it be better to torrent them to the school server?
17:45 < hhardy> lfaraone: one of your solutions looking for a problem has found a problem, grats :)
17:45 < dogi> +1
17:46 < lfaraone> hhardy: welcome! :)
17:46 < cjl> The thing about it is tha these are not "real hits".  They are just exploratory clicks on the Actvbities link on Browse front page (for hte most part.
17:46 < lfaraone> adric: maybe... but that'd be a bittorrent extention, after we get the protocol implemented.
17:47 < hhardy> lfaraone: ACTION ITEM: publish a plan for including torrent capability in browse, and a backend to provide large images via torrent
17:47 < lfaraone> cjl: that should be fixed with XS + squid
17:47 < hhardy> that ok?
17:47 < cjl> Other than the "bringing server ot it's knees" thing, this not a "real" problem.
17:47 < lfaraone> hhardy: yes. but it'll have to wait a few days.
17:47 < hhardy> in 2 weeks?
17:48 < lfaraone> hhardy: yes, I should be able to have it ready by then.
17:48 < hhardy> yeah well the pringing the server to its knees is not good on pedal because that has our email, mailman and most of the websites
17:48 < cjl> hhardy: granted
17:49 < lfaraone> hhardy: ah, _that's_ why my email has been so late in arriving through l.o... :)
17:49 < hhardy> send me full headers of "slow" email please
17:49 < hhardy> I have the postfix set to notify me of delayed mail
17:49 < hhardy> it hardly ever fesses up
17:49 < hhardy> but I still hear scattered reports of slow mail
17:49 < lfaraone> hhardy: sure. sometimes I get mail a day later. 
17:50 < cjl> I'm just saying that the emphasis on tech/infrastructure solutions should not assume that these really "matter" as most are just random click-throughs..
17:50 < lfaraone> hhardy: but it's been better recently.
17:50 < hhardy> uruguay matters and to an extent, g1g1 matters
17:50 < hhardy> oh in terms of mail
17:50 < lfaraone> hhardy: usually it's a short while.
17:50 < hhardy> well it is a concern for some of the senior management here so if we can identify the issue its a priority for me
17:51 < lfaraone> hhardy: mail, or wiki?
17:51 < cjl> hhardy: Of course UY matters, but I am inferring from the data that these are random clickthroughs, not real intentional requests for new activity page views.
17:51 < hhardy> both but I meant mail
17:51 < hhardy> sorry offtopic
17:51 < hhardy> as far as uruguay, seems like kids exploring from the default links on browse
17:52 < hhardy> btw build 8.2 is ready so please test it
17:52 < cjl> hhardy: exactly
17:52 < lfaraone> hhardy: we can make it so that a bit of url-fu is passed along to the wiki when it is from the browse page...
17:52 < lfaraone> hhardy: that sends it off to a cache server...
17:52 < hhardy> that is possible, and interesting
17:52 < cjl> I'd have to look, but I bet "Activities" got translated as something intersting to kids like "Juegos".
17:53 < hhardy> browse could report the sku of the machine for instance, but people might think that is spyware-like
17:53 < lfaraone> hhardy: no, that's creepy.
17:54 < hhardy> yeah its a fine line
17:54 < hhardy> if we are upfront about what we are doing we might get beaten up but it reduces the creep factor
17:54 < lfaraone> hhardy: I mean something like http://wiki.laptop.org/go/Foo?source=homepage&language=es_AR
17:54 < adric> http_referrer, aye
17:55 < cjl> DOes this work for you guys
17:55 < cjl> https://www.google.com/analytics/reporting/map_detail?id=10025922&pdr=20080906-20081006&cmp=average&rpt=GeoMapReport&d1=UY&segkey=request_uri_1&mdet=COUNTRY&seg=1#lts=1223415858997
17:55 < lfaraone> adric: does that work for references from file:/// pseuo-protocol?
17:55 < cjl> That shows Uruguay broken down by landing page.
17:55 < adric> lfaraone: could, sure
17:56 < cjl> 238K of total 263K hits land on Activities.
17:56 < lfaraone> adric: google doesn't seem to have that data: https://www.google.com/analytics/reporting/referring_sources?id=10025922&pdr=20081007-20081007&cmp=average&trows=10&gdfmt=nth_week#lts=1223416527797
17:56 < kimquirk> hhardy: did we discuss vservers on Grinch and Solar for RT instances?
17:57 < hhardy> no thats later in agenda lets move on
17:57 < hhardy> lfaraone: do you want to say anything more abotu single-signon?
17:57 -!- adric_ [~adric@adsl-160-67-156.asm.bellsouth.net] has joined #olpc-admin
17:58 < hhardy> assume he is tied up
17:58 < cjl> One last thing, remember how important Activities page is to auto-update feature in any plans.

SAGE

17:58 < hhardy> send in your gpg signed agreement to follow the SAGE ethics code pls
17:58 < hhardy> prefer to the olpc-sysadmin list, or to me if privacy/security is an issue
17:59 < adric_> oops. here
17:59 < lfaraone> hhardy: back.
17:59 < hhardy> the current situation of using a publically editable wiki page to provide the activities for the updater is... rube goldbergish?
17:59 < hhardy> and not so secure?
17:59 < dogi> hhardy: am i ok?
17:59 < lfaraone> hhardy: not really, other than I still think it's useful, but whatever.
17:59 < lfaraone> hhardy: yes, it is very scary.
17:59 < lfaraone> hhardy: anybody can change a URL.
18:00 < lfaraone> hhardy: Has anybody worked on converting mozilla Addons' site?
18:00 < hhardy> lfaraone: not so far as I know
18:00 < hhardy> dogi: are you ok? as far as I know...
18:00 < cjl> hhardy: no argument there, just point out that it needs to be factored in (as reality for nowa).
18:00 < dogi> thx
18:00 < dogi> with the sage signing stuff
18:01 < cjl> hhardy: Is pdf scan of wet-signed good for you?  Used PGP in '92 when I was on the cypherpunks list, but I've been all X-509 since.
18:01 < hhardy> ah ok thanks
18:01 < hhardy> yes dogi is: OK
18:01 < hhardy> sure
18:01 < hhardy> I am encouraging the use of gpg, we can all put out keys on the MIT server
18:01 < hhardy> http://pgp.mit.edu
18:02 < adric_> cjl; recent GPG can use x509 ;)
18:02 < hhardy> s/out/our/
18:02  * lfaraone is afk.
18:02 < hhardy> for cryptographic reasons, it is good to use a different key for signing and for encrypting, btw
18:03 < cjl> Will look int ogenerating PKI pair again, old ones long gone into bit rot.
18:03 < hhardy> anyway we can have a signing party some time and sign each others keys
18:03 < dogi> :)
18:04 < hhardy> I was thinking if we can build some momentum it might be nice to do a face to face maybe in the winter or so
18:04 < cjl> gnu{-: used to host the c-punks list on toad.com IIRC.
18:04 < hhardy> how many people would be willing/able to come up to Boston sometime?
18:04 -!- adric [~adric@adsl-158-6-244.asm.bellsouth.net] has quit [Ping timeout: 480 seconds]
18:05 -!- adric_ [~adric@adsl-160-67-156.asm.bellsouth.net] has quit [Quit: *bampf*]
18:05 -!- adric [~adric@adsl-160-67-156.asm.bellsouth.net] has joined #olpc-admin
18:05 < hhardy> http://www.cypherpunks.to/list/
18:05 < hhardy> adric: you back?
18:05 < adric> Still thinking of FudCon, in Jan ?
18:06 < hhardy> thats a possible
18:06 < cjl> hhardy: some possibility of my drifting up to Boston (from NJ), but it is likely to be on short notice and at the mercy of someone else's schedule.
18:06  * adric is a Thawte and CACert Notary, if that's useful. 
18:06 < hhardy> kimquirk was reminding me we need to set up a copy of the rt instance for adric's team to use as a sandbox
18:06 < hhardy> and I need to set up one for internal use on... an itnernal machine
18:07 < kimquirk> m_stone: are you the vserver expert?
18:07 < hhardy> mstone: ping
18:07 < hhardy> m_stone: ping
18:07 < kimquirk> m_stone: we are thinking of RT instances that match the one we have today... a vserver on solar (I believe
18:08 < kimquirk> any other vserver experts?
18:08 < kimquirk> hhardy: I'll send out an email
18:08 < hhardy> http://linux-vserver.org/HowtoCloneVserver
18:09 < hhardy> I've invited Ed and the engineering folks many times, I need to bake cookies or something to encourage them
18:09 < hhardy> to this mtng
18:09 < hhardy> actually Ed I have said it maybe 1 or 2 times
18:10 < kimquirk> he said he will try to make it next week. he was feeling ill and left early
18:10 < hhardy> maybe with 8.2 ou the door this will be more possible 
18:10 < hhardy> ok
18:10 < kimquirk> i think so.
18:10 < hhardy> adric do you have time this week to take a crack at cloinigng the rt vserver with me?
18:11 < adric> hhardy: I think so, yeah .. daytime or nighttime ?
18:11 < hhardy> morning, night or weekend
18:11 < hhardy> afternoons are my rush time
18:11 < hhardy> we have monday off
18:12 < hhardy> I'm officially not working but might to some admin for fun lol
18:12 < adric> heh, ok. I should be in an office tomorrow day, and then available later
18:12 < hhardy> adric email me or grab me on this chan and let's try to get this done
18:13 < hhardy> ok maybe I will timeshift tomorrow and take some evening time
18:13 < adric> ACTION ITEM: clone rt on solar.

Big Sister

18:13 < hhardy> calling again for a Big Sister maintainer
18:13 < hhardy> ok we will leave that one open
18:14 < hhardy> any new business?
18:14 < hhardy> or anything missed?
18:14 < cjl> any further discussion about spam?
18:14 < adric> I meant to mail out some thoughts on RT and email and spam. Will try and do that this week
18:14 < hhardy> mb has been working on the issue and specifically blacklisting the worst offenders
18:15 < cjl> I've personally rejected over 2000 spam messages from RT, thus the curiousity.
18:15 < hhardy> I gave him an rt account so I can add him to tickets that are of interest to him
18:15 < hhardy> please dont you all start assigning him things he might be annoyed
18:15 < adric> The grail is  a system that feedsback spam tagging from RT back to the cuda.
18:15 < hhardy> yes
18:16 < hhardy> something automatic would be a godsend
18:16 < cjl> From what I've seen nothing with the [BULK] tag is ever worth keeping.
18:16 < adric> nt entirely true, sadly
18:16 < adric> I have seen a few false negs.
18:17 < cjl> That's the challenge with hand te3aking the stringencies.
18:17 < hhardy> I got a great spam from NASA Tech Briefs to which I do subscribe
18:17 < adric> Anyway, two-pronged, or three for now, with unification the goal?
18:17 < adric> I will try and exlain in mail / wiki this week
18:17 < hhardy> stop spam, by any means necessary :) (almost any means)
18:17 < adric> Wait, can we kill spammers?
18:17  * cjl still holding up the manual despamming with a boyscout knife end (for now)
18:17 < hhardy> *almost lol
18:18 < adric> Thanks for that cjl, I get in a little oof that with my spork some days.
18:18 < hhardy> best spam ever: "Don't miss "The Power of Beryllium Materials" complimentary webinar"
18:18 < adric> wow.
18:19 < adric> you shouls send tht on to the writers of The Big Bag Theory sitcom
18:19 < adric> bong
18:19 < adric> bang
18:19 < cjl> There has been a lot ot totally random Russian spam recently.  (I ususally check with google)
18:19 < hhardy> lol... moving right along... anything further?
18:19 < adric> there's always been a lot of russian business offers ;)
18:19 < cjl> no, reaallllly random
18:20 < hhardy> That is all...

Afterhour

18:20 < hhardy> ---END---
18:20 < hhardy> thanks all for coming
18:20 < dogi> ciao
18:20 < hhardy> I am going to see about restoring my ops on this chan, is anyone who will watch the chan who wants ops?
18:21 < adric> er, okay?
18:21 < adric> If it will help. I try to keep a window open in here.
18:21 < hhardy> people try to beat baysian filters using markov chaining... kind of funny
18:21 < hhardy> random arms race
18:21 < adric> hehe
18:22 < cjl> Actually, on the distinction between spam and "unsolicited commercial e-mail" that is something we might want to think about.  There is spam that you definitely want feeding into whatever black list or heuristic you use and then there is stuff that falls a little short of needing that kind of hammer.
18:24 < adric> well ... If management wants to play it that way, we can try.
18:24 < hhardy> there's never going to be a clear line between "I never want to see this", "I might want to see this" and "I always want to see this" as it is a continuum
18:24 < adric> Which categoriy is solicitation for donatio of laptops to $worthycause ?
18:24 < adric> :)
18:24 < cjl> It is really hard to tell if all the invitations to join various "friending" opportunities are people who are just clueless, or if it is some social engineering scheme with link traps
18:25 < hhardy> yea
18:25 < adric> agree
18:25 < cjl> I have replyied toi a few (legit) newsletters tha just should be going to RT handled address wit hUNSUBSCRIBE msgs.
18:26 < hhardy> I generally try to follow Postel's Law
18:26 < adric> Postel's law, or incoming mail .. hmm ...
18:27 < adric> Bah, typing. SO is waiting, i shall have to dicuss this fascinating point of security philosophy with you gentlemen (and kim) later :D
18:27 < hhardy> RFC 793:   TCP implementations will follow a general principle of robustness:  be conservative in what you do, be liberal in what you accept from others.
18:27 < hhardy> ok cul8r
18:28 < adric> accept sure, and immediatey roundfile, maybe ..
18:28 < adric> night !
18:28 < hhardy> including significant others :)
18:28 -!- adric [~adric@adsl-160-67-156.asm.bellsouth.net] has quit [Quit: adric]
18:29 < hhardy> heading out night all
18:37 -!- mode/#olpc-admin [+o hhardy] by ChanServ
18:39 -!- cjl [0c2c32f8@webchat.mibbit.com] has quit [Quit: http://www.mibbit.com ajax IRC Client]
18:43 -!- kimquirk [~kimquirk@can-olpc-nat.media.mit.edu] has left #olpc-admin []
18:44 -!- isforinsects [~isforinse@140.247.243.123] has quit [Ping timeout: 480 seconds]
18:47 <@hhardy> adric: you will have to register on oftc before I can give you master on the chan
18:49 <@hhardy> adric: /msg Nickserv register <password> <email>
18:57 < dogi> hhardy: can u send me to log of this meeting ... my chatclient was not in logging mode :P
19:03 -!- dogi [~dogi@dsl092-078-228.bos1.dsl.speakeasy.net] has quit [Quit: Leaving]
19:03 -!- dogi [~dogi@dsl092-078-228.bos1.dsl.speakeasy.net] has joined #olpc-admin
19:21 <@hhardy> dogi ok