MoinMoin: MoinMoinChat/Logs/moin-dev/2008-02-28

2008-02-28T00:03:25  <ThomasWaldmann> look now
2008-02-28T00:03:46  <ThomasWaldmann> XXX is todo
2008-02-28T00:05:26  <xorAxAx> i will have a look at it at the weekend, after my last exam for this month :)
2008-02-28T00:10:50  <ThomasWaldmann> heh, "for this month"
2008-02-28T00:16:07  <dreimark> good night
2008-02-28T00:16:47  <ThomasWaldmann> gn dreimark
2008-02-28T00:31:54  <grzywacz> night
2008-02-28T08:19:37  <cb22> xorAxAx: dates are going to be fun
2008-02-28T08:26:16  * ThomasWaldmann plays with varnish
2008-02-28T08:40:57  <cb22> ThomasWaldmann: it rocks
2008-02-28T09:08:33  *** irc.freenode.net sets mode: +o ChanServ
2008-02-28T09:15:07  <ThomasWaldmann> cb22: i guess it does, when content is better cacheable than moin
2008-02-28T09:15:57  <xorAxAx> cb22: hehe, yep :)
2008-02-28T09:24:29  <ThomasWaldmann> cb22: do you use varnish?
2008-02-28T09:35:19  <TheSheep> displaying a logged in user name of every page makes cacing only available per-user
2008-02-28T09:35:41  <TheSheep> good thing it doesn't display the current time like phpbb
2008-02-28T09:35:47  <xorAxAx> ?
2008-02-28T09:35:58  <xorAxAx> ah, http caching, yes
2008-02-28T09:36:16  <xorAxAx> TheSheep: well, usually caching the data for anonymous users is enough to ensure performance
2008-02-28T09:36:31  <xorAxAx> as long as the caches can derive from the headers that logged in requests may not be cached
2008-02-28T09:36:33  <TheSheep> depends on how your users usually use the site
2008-02-28T09:36:50  <TheSheep> if you require login, for example...
2008-02-28T09:37:29  <TheSheep> we have a huge social networking site here that grew explosively over the last year. they are completely helpless
2008-02-28T09:37:35  <TheSheep> and hapless
2008-02-28T09:37:44  <TheSheep> but they just bough larger servers
2008-02-28T09:37:49  <TheSheep> bought
2008-02-28T09:37:57  <xorAxAx> well, yes, then you need different ways to scale :)
2008-02-28T09:42:06  <dreimark> TheSheep: do you like to become a mentor for gsoc?
2008-02-28T09:42:48  <TheSheep> dreimark: yes
2008-02-28T09:43:20  <dreimark> please add your name then to http://moinmo.in/GoogleSoc2008
2008-02-28T09:44:04  <dreimark> TheSheep: fine
2008-02-28T09:49:55  <TheSheep> I wonder if using _-prefixd functions from python stdlib would be wise :)
2008-02-28T09:50:29  <TheSheep> well, if it breaks I'll just write my own
2008-02-28T10:45:23  <grzybacz> moin
2008-02-28T10:48:47  <TheSheep> badger badger badger badger badger badger MUSHROOM! MUSHROOM!
2008-02-28T10:56:04  <grzybacz> What did you smoke today? :-)
2008-02-28T10:56:08  <grzybacz> SNAAAKE!
2008-02-28T14:31:51  <dreimark> TheSheep: http://www.albinoblacksheep.com/flash/badgers
2008-02-28T14:39:15  * cb22 trips
2008-02-28T17:53:23  * cb22 fixes traceback on unknown action
2008-02-28T17:53:40  <cb22> ThomasWaldmann: btw yes, i use varnish
2008-02-28T17:53:51  <cb22> it speeds things up considerably
2008-02-28T17:54:04  <ThomasWaldmann> yeah, except moin :)
2008-02-28T17:54:23  <cb22> hehe, guess i'm used to slow php apps :p
2008-02-28T17:54:40  <johill> I find that moin's rendering speed depends on the speed that the css is loaded etc
2008-02-28T17:54:47  <johill> I had much improvement with lighttpd over apache
2008-02-28T17:54:57  <ThomasWaldmann> the problem is that wiki's don't have much static content
2008-02-28T17:55:08  <ThomasWaldmann> or cacheable content
2008-02-28T17:55:15  <johill> cacheable
2008-02-28T17:55:17  <johill> that reminds me
2008-02-28T17:55:25  <johill> I fixed the failing anonymous session tests
2008-02-28T17:56:26  <CIA-38> Johannes Berg <johannes AT sipsolutions DOT net> default * 3147:47b20bf81097 1.7/MoinMoin/auth/_tests/test_auth.py: fix and comment failing auth tests
2008-02-28T17:56:31  <cb22> johill: yep, lighttpd rocks
2008-02-28T17:56:52  <cb22> or pebbles, when you look at its size :)
2008-02-28T17:56:56  <CIA-38> Johannes Berg <johannes AT sipsolutions DOT net> default * 3148:e1fe8dd52b83 1.7/MoinMoin/action/serveopenid.py: pep8-fix openid provider
2008-02-28T17:57:48  <johill> ThomasWaldmann: the problem with the trail test is that the first request isn't written to trail
2008-02-28T17:58:15  <johill> ThomasWaldmann: that's to avoid creating anonymous sessions when the requester didn't send us a cookie, because otherwise each wget request etc. would create a session
2008-02-28T17:58:41  <johill> that hits logged in too though, currently
2008-02-28T18:02:58  <xorAxAx> i dont see why wget is a reason
2008-02-28T18:03:32  <johill> but you do understand the problem?
2008-02-28T18:04:00  <xorAxAx> yeah, but i dont see how that is a relevant mitigating factor
2008-02-28T18:04:16  <xorAxAx> because wget will give you the cookie it gets back on first request, right?
2008-02-28T18:04:26  <xorAxAx> if you use it to mirror etc.
2008-02-28T18:04:33  <johill> oh in that case it'll get a trail too
2008-02-28T18:04:44  <xorAxAx> yes
2008-02-28T18:04:46  <johill> if it hands back the cookie
2008-02-28T18:04:50  <johill> and that's fine
2008-02-28T18:05:04  <johill> I just want to avoid generating a new session file for each request if the user agent doesn't hand back cookies
2008-02-28T18:05:22  <xorAxAx> it would be interesting to see what the request ratio no cookie/a cookie is
2008-02-28T18:05:23  <xorAxAx> hmm
2008-02-28T18:05:37  <johill> of course you can trivially code an app that does a non-cookie/cooke request storm
2008-02-28T18:05:43  <xorAxAx> well, wget does. browsers with disabled cookies dont
2008-02-28T18:05:48  <johill> i.e. do each request twice once with and once w/o cookies
2008-02-28T18:05:54  <johill> that'll still create lots of files
2008-02-28T18:05:59  <johill> but browsers with disabled cookies won't
2008-02-28T18:06:05  <johill> so that's the rationale
2008-02-28T18:06:23  <xorAxAx> i have never heard that other apps do that
2008-02-28T18:06:35  <johill> do what?
2008-02-28T18:06:50  <xorAxAx> defer cookie generation to the second request in a session
2008-02-28T18:06:58  <johill> I don't defer cookie generation
2008-02-28T18:07:01  <xorAxAx> or session generation
2008-02-28T18:07:02  <xorAxAx> yes
2008-02-28T18:07:12  <johill> I just don't save anything into the session
2008-02-28T18:07:22  <johill> if some other code actually does save things into the session it *will* be saved
2008-02-28T18:07:32  <xorAxAx> i rather think that sane on disk data structures and house keeping is worth more than these kind of fixes that impact usability a bit
2008-02-28T18:07:36  <johill> it's just that the trail isn't, as an optimisation against having that many files
2008-02-28T18:10:16  <johill> I dunno. session file storage is cheap too
2008-02-28T18:10:24  <johill> after all, that's what filesystems are good for, storing files
2008-02-28T18:10:29  <johill> except maybe FAT or so
2008-02-28T18:11:37  <xorAxAx> johill: many file systems slow down starting with a certain number of files in the dir
2008-02-28T18:11:46  <xorAxAx> ext3 does aswell if dir indexes are disabled
2008-02-28T18:12:01  <xorAxAx> thats why such structures are often fanned out hash tries
2008-02-28T18:12:22  <johill> yeah but dir indexing is how old?
2008-02-28T18:12:47  <xorAxAx> no idea :) you know, think about the redhat admins using python 1.5 :)
2008-02-28T18:12:50  <johill> I don't really believe in doing things in an app that the underlying code can handle
2008-02-28T18:13:06  <ThomasWaldmann> johill: thanks for fixing
2008-02-28T18:13:21  <xorAxAx> johill: hmm?
2008-02-28T18:13:51  <johill> xorAxAx: like writing a session storage database just because your filessystem can't handle a few thousand files
2008-02-28T18:14:19  <xorAxAx> cb22: "Also, I originally entered these dates into the calendar as
2008-02-28T18:14:20  <xorAxAx> unavailable, but I can't see them unless I switch to week view."
2008-02-28T18:14:33  <xorAxAx> cb22: does that mean leslie incorrectly determined that date to be free?
2008-02-28T18:14:41  <xorAxAx> cb22: or that the calendar app is broken?
2008-02-28T18:14:51  <xorAxAx> johill: no, db is fine
2008-02-28T18:14:53  <xorAxAx> umm, fs
2008-02-28T18:15:02  <johill> anyway, feel free to remove the code that doesn't store the trail on the first request, I just feel that is a worthy optimistion so we don't hit the fs for every non-cookie request
2008-02-28T18:15:03  <xorAxAx> i mean, you can just do better than a single dir
2008-02-28T18:15:07  <cb22> xorAxAx: i think google calendar is tripping because i made the dates reoccuring
2008-02-28T18:15:35  <johill> many apps just store the trail in another cookie, maybe that's a solution ;)
2008-02-28T18:15:42  <xorAxAx> johill: i would first seek info about the ratio of requests that are going through that if branch and those which dont :)
2008-02-28T18:15:56  <xorAxAx> cb22: ah
2008-02-28T18:16:04  <xorAxAx> cb22: and she probably used list view?
2008-02-28T18:16:09  <johill> I guess I could log that on linuxwireless.org, it's a failry high traffic site
2008-02-28T18:18:06  <johill> xorAxAx: of course, I would htink that spiders like google don't send a cookie either
2008-02-28T18:22:05  <dreimark> johill: do you like to mentor at gsoc ?
2008-02-28T18:23:52  <johill> I suppose I could, but I won't fight for it
2008-02-28T18:23:59  <johill> I mean, if somebody else wants to ... :)
2008-02-28T18:24:32  <dreimark> ok, but currently we don't know how much students we can get
2008-02-28T18:24:38  <dreimark> but we know one for one mentor and it increases every year
2008-02-28T18:25:18  <dreimark> so please add your name to http://moinmo.in/GoogleSoc2008
2008-02-28T18:25:37  <johill> right
2008-02-28T18:31:17  <dreimark> bbl
2008-02-28T20:11:01  <xorAxAx> cb22: would be nice if you could make it
2008-02-28T20:37:28  * ThomasWaldmann continues on AttachFile
2008-02-28T20:45:45  <cb22> xorAxAx: indeed :) just been emailing with leslie, looks like they could pay for my trip to italy o-O
2008-02-28T20:46:31  <cb22> (i'd cancel the air miles ticket, get a full refund and then do cape town -> italy -> san fran -> cape town
2008-02-28T20:47:04  <xorAxAx> cool
2008-02-28T20:47:17  <xorAxAx> cb22: well, its certainly easing leslies scheduling burdens :)
2008-02-28T20:55:49  <cb22> ThomasWaldmann: ping
2008-02-28T21:49:14  <CIA-38> Thomas Waldmann <tw AT waldmann-edv DOT de> default * 3149:6749e003d7e2 1.7/MoinMoin/ (4 files in 4 dirs): remove attachments direct serving (cfg.attachments), refactor AttachFile to use formatter for link generation
2008-02-28T21:49:15  <CIA-38> Thomas Waldmann <tw AT waldmann-edv DOT de> default * 3150:8ae94675b9f9 1.7/MoinMoin/ (3 files in 3 dirs): AttachFile: some cosmetical source changes
2008-02-28T21:49:19  <CIA-38> Thomas Waldmann <tw AT waldmann-edv DOT de> default * 3151:863d90c05507 1.7/MoinMoin/action/AttachFile.py: AttachFile: more refactorings
2008-02-28T22:26:15  <CIA-38> Thomas Waldmann <tw AT waldmann-edv DOT de> default * 3152:719c5bf80150 1.7/MoinMoin/action/AttachFile.py: AttachFile: move ACL processing to handlers, simplify