[12:42:21] hi can someone help me troubleshoot my svn access [13:40:26] jeremyb: you around ? [14:54:48] OrenBochman: hi [14:54:57] I will be happy to help you [14:55:36] hi can you check is my subversion certificate is enabled [14:55:41] I cannot loggin [14:55:50] or rather checkout [14:55:58] ok, what does it say? [14:56:07] you are oren? [14:56:11] yes [14:56:17] what client you have? [14:57:11] disconnected: No supported authentication methods available (server sent public key) [14:57:20] tortoisesvn [14:57:25] also eclipse [14:57:30] ok, click checkout and type to URL [14:57:36] svn+ssh://oren@svn.wikimedia.org/svroot/mediawiki/trunk/tools/nparser [14:57:41] try to click ok [14:58:11] it's possible that you don't have your key correctly imported [14:58:31] do you have any other svn client available? I never tried to do this on windows [14:58:45] eclipse subversive [14:58:55] but the error you say happens when you don't have ssh private key available [14:59:09] I see [14:59:39] http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CCgQFjAB&url=http%3A%2F%2Fwww.vectorns.com%2Fblog%2F11-tortoisesvn-over-ssh-on-windows-via-putty&ei=0Iv4Trj6M6jj4QSZpLj4Dg&usg=AFQjCNFzATpUdnv-CsQAOMkOpWu1Pz8eLQ [14:59:47] oh [14:59:52] www.vectorns.com/.../11-tortoisesvn-over-ssh-on-windows-via-putty [14:59:56] that's what I meant [15:00:03] google fucked up link [15:03:18] there is something there I have not tried [15:03:26] saving the putty session [15:03:35] ok [15:03:52] what is nparser [15:04:26] that's a small directory we can try a checkout there :) [15:04:34] I didn't know any other small directory to test [15:04:57] it contains 3 files [15:07:34] same issue [15:07:41] unfortunatly i have to go now [15:07:43] ttl [15:07:44] ok [15:45:43] petan: I'm here now if you still want me. [15:45:56] I wanted to discuss the protocol we talked about :D [15:46:20] I want to start a page on mww where we could write down specifications and update docs [15:46:52] people on wikitech said we could use XMPP instead of irc [15:47:14] indeed, though it depends what we would be aiming for. [15:47:52] I would like to create open communication protocol with easy integration to existing tools [15:50:58] Yeah but there are 2 sides to this, either a) having edits feeds like we do currently or b) more custom inter-project data using a standard format. For example currently cbng parses the talk page to see how many warnings+what level of warning a user has because the nearest to a standard is the templates. There is no way to query that. [15:51:12] that could be shared [15:51:32] there would be a query to communication channel if client have information about this talk page, if some client would have it it would share it with others [15:51:38] So the question becomes do we want a better, more reliable edit feed in a better more standard format where each tool parses it and decides if it's needed or a format where we split it down more into actions so an edfit that is give user x warning xx level xx is a message in it's self. [15:51:47] so once a client parse number of warnings it could share it [15:52:14] no we don't want to change the current feed we want to create entirely new feed [15:52:24] which would be used together with existing [15:52:42] so that tool would be using the current feed + the new one so it would work much faster than now [15:53:15] for instance cluebot would see edit with 40% probability instead of reverting it would send it to feed [15:53:20] so that other editors could review it by hand [15:53:40] if it was vandalism they would send feedback to cluebot so it could update its db for future [15:54:02] Hmm, from a anti vandal point of view having the current edit feed in a format where you can get older entries makes sense as a bot restart = hundreds of edits misses. [15:54:32] that could be also fixed using new one [15:54:37] If we take an example with cbng for anything within a threashold just under the revert threashold which is like maybe 20% of edits that makes sense. [15:54:48] indede [15:54:59] As far as adding it into a central review collection like cbng uses for it's ann we have another issue of data storage and integrity. [15:55:34] Because a big issue is getting bad reviewed data into the dataset for cbng, which is why all review users are manually reviewed and we have certain checks to ensure there is a large majority in agreement. [15:55:37] yeh also it would avoid double checks of same edits, because if cluebot or used decided edit is ok, there is no need for review by others [15:55:45] so that antivandalism work would be way faster [15:55:55] * user [15:55:59] not used :P [15:56:49] Indeed, which is what annoys most people - cbng's speed. [15:57:40] Ideally we where looking to scrap the cbng review interface and integrate it into stiki/huggle so people would check edits as normal and they would either be a) live edits that are not yet reverted or b) non-live edits that just get added/updated in the dataset which we need multiple reviews for. [15:57:42] I heard that it's also annoying to review same edits which are ok [15:57:50] maybe even more than reverting same edit in same time [15:59:08] my idea is that huggle would make same queue but if cluebot sent that it just skipped x% vandalism edit, it would put edit to top [15:59:19] so that user could review it for cluebot and send feedback back to it [15:59:32] also if another user would revert edit, huggle would remove it from queue [15:59:39] same with edits which are marked ok [15:59:45] Hmmm [15:59:52] This raises another question [15:59:56] so that users would need to review only edits others are not sure about or no one reviewed [16:00:08] instead of all edits even reverted [16:01:46] Currently each tool has their own backends, cbng's uses an ann, huggle uses ?, stiki has about 5. So from a messaging point of view sharing data such as edit xxxx has been reverted, don't bother showing it to anyone else makes sense. But from a point of creating a huge store of edits it's poses a lot of challenges. [16:02:21] huggle doesn't have a backend because for performance reasons it doesn't preparse more edits than 5 or 10 [16:02:25] An approach of having WUI/GUI tools talking to a backend that does the ann stuff so feed <> bots <> queue <> backend <> ui would make more sense but take a lot more inter-project work because currently there are like 3? popluar human based tools and cbg [16:02:29] it only filters the queue [16:02:32] ah [16:02:38] so that you can pick if you want to see only ip edits etc. [16:02:48] It just pikc up everything flagged by edit filters? [16:02:57] that can be also filtered [16:03:04] you configure what you want to display [16:03:08] you can even display all edits [16:03:15] I never got huggle working properly so have never used it tbh. [16:03:35] default is non whitelisted + flagged on top [16:03:51] it's very simple but still effective pretty much [16:05:33] Does huggle talk to the mediawiki api directly then and just abstract over that? [16:05:44] actually this protocol isn't a complete ann like backend, just a protocol so that clients can share data [16:05:57] yes it uses api [16:06:27] it download a list of recent changes and parse them according to filters [16:06:36] then it start precaching the edits as they are in queue [16:06:50] Basically what we need at a very high level is a secure way to share data, one that can't have people insert bad data into (which is quite tricky if they just talk to the api directly and each client will talk to the queue seperatly). [16:07:24] hm, that's true probably there should be some authentication method like oauth [16:07:32] or something like that [16:07:48] Which would mean on an even the tool would store an action into the queue to say done xx on xx and the other tools from that would know that edit has been done allready which from a ui point of view means not showing, from cluebots it means no spending 40seconds waiting on the api. [16:08:28] that would make it a lot faster then [16:08:41] It would be kinda of nice if there was a way to use mw logins better... because the current way works fine for storing it in a session state but sucks from checking it and using it in a 3rd party app for security reasons. [16:09:02] It would also mean we can track back things to a certain user easier. [16:09:07] I think having an open id provider for SUL is a good idea [16:09:17] like you could use SUL as open id [16:09:22] This is why the cbng report interface half sucks.... [16:09:29] We can't validate securly who reports an edit. [16:09:34] all tools suck thanks to that [16:09:47] http://www.mediawiki.org/wiki/OAuth [16:10:13] That would be nice [16:10:17] definitely [16:10:24] What does the MWF use in the backend anyway? LDAP or some bespoke crud? [16:10:31] for what? [16:10:35] (for inter-site sso/login stuff) [16:10:40] ldap [16:11:33] Btw did you see the link I sent in tec? [16:11:38] Like a few days ago [16:15:02] probably not [16:15:03] where [16:16:02] ^demon: \o, [16:17:24] <^demon> petan: Howdy. [16:18:48] Ah it dm then [16:26:34] Thinking about this more it might be wise to take more of a queue based approach. Say for example if we broadcast out a user warning level and there was a message still pending delivery rather than spam a client with hundreds of messages that are outdated (say if they are over a REALLY slow link) it would make more sense to drop the first message out the queue and replace it with the second. [16:27:16] There would also probably have to be a way for a client to specify what type of events it wants to see, or there would be the same properly. Scores etc probably don't need to be known by users, but reverts do for example. [16:27:54] * Damianz goes and noms tea [17:10:42] that's up to tool what info it would use [17:11:12] more I think of that I think it would be best if we created own server for that, not using any of existing protocols [17:11:26] because implementing oauth to anything is probably hard [17:14:20] OrenBochman: hi [17:14:32] OrenBochman: i'll be back in half an hour [17:22:48] Indeed.... [17:23:02] Though then we have an issue of message routing and redundancy which are painful. [17:23:31] If we could take a system like Ryan was talking about, write an authentication gateway for it (so clients can use session tokens after which are generated from the auth server) that would be win. [17:23:43] no, if you get same report for same edit, it should be same report or it should be dropped [17:23:51] in case it's same it's ok [17:23:59] in case it's not same, it's weird and ignored [17:24:24] that's what I meant we could create a c++ server app [17:24:24] I mean from the point of view if I run around the dc with a hammer and start having a psycho episode with the switches [17:24:28] :P [17:24:49] heh [17:25:09] But in theory a mesh style messaging framework basterdised.....errr bespokly written for our needs would work [17:26:41] exactly [17:28:05] Now if only I'd bothered to learn C++ over perl ;) [17:28:40] cb is written in c++ [17:28:43] you should know it heh [17:29:02] Oh I've edited bits of cb's core, doesn't mean I know c++ [17:29:02] :P [17:29:08] ah [17:29:30] everyone know the exotic languages like perl or python and no one knows the english of programming :D [17:29:38] I've written stuff in C that's running in production but I've never learned C, I just get told off by people who know C for my bad code xD [17:29:46] Oh shh [17:29:49] heh [17:29:58] The amount of stick I get for choosing python over learning assembly [17:30:23] cool applications are written in c, like linux :) [17:30:27] C just drives me mad, it's so slow to develop in. Like 500 limes to get something basic to compile is =/ But yeah C++ is a little nicer. [17:30:29] XD [17:30:41] I don't think some of the linux kernel counts as C ;) [17:30:46] heh [17:30:49] why not [17:31:12] c support inline asm :P [17:31:18] that's cool [17:31:19] C++ too [17:31:22] I know [17:31:40] * petan <3 c and c++ [17:31:54] I like c but for little things :P [17:32:13] cbng's core is pretty basic C++ tbh, apart from libconfig which I've never understood properly. [17:32:34] I like wxwidgets, that' cross platform c++ library [17:32:44] so you can create form applications for all platforms [17:32:54] I chose c# over c++ for a little gui thing I wrote. [17:32:59] omg you're developing GUI [17:33:11] MaxSem: sometimes it's useful to have a gui :P [17:33:21] Damianz: huggle is in c# [17:33:21] blehhh [17:33:28] Actually for one gui I wrote a java wrapper then packaged up Jython as a jar *yuck* [17:33:32] MaxSem: wanna terminal huggle? [17:33:43] TUI huggle would be win [17:33:51] XD [17:33:54] GUI is not bad [17:34:16] GUI in C++ is masochism [17:34:39] not if you have a good libraries for that [17:34:51] wxwidgets offers possibilites of mono [17:35:15] wxWidgets > gtk? [17:35:20] I think so [17:35:34] meh, I escaped from C++ world and live fine without it [17:35:35] it's not just a framework for gui [17:35:38] it's more like .net [17:35:45] Ah [17:35:51] .net is pretty awesome for form stuff [17:36:07] but it's open source and for c++, .net in c++ suck [17:36:14] you need to use cli [17:36:31] I've been doing simple stuff in cli c++ and it was terrible [17:36:49] it's like a microsoft standart which even microsoft support badly [17:36:54] I use to "maintain" (read as fix it when it broke) a .net based vb gui thing that was just a huge interface to mysql. [17:37:26] vb is sort of like c# just even simpler and not portable [17:37:33] vb is horid [17:37:36] completely different syntax [17:37:38] c# is like nice java to me [17:37:45] Java drives me potty [17:38:01] c# is like java with c++ features :) [17:38:17] Now if only mono was a tiny bit better. [17:38:19] I don't know much java, but I've read some comparison on wikipedia [17:38:31] it was clear that c# has all what java is missing [17:38:54] mono isn't bad at least if you develop stuff in that, it works [17:38:57] java is just meh, I keep thinking about improving cbng's review interface but it's all java :( [17:39:05] problem is to get .net code work on mono sometimes [17:39:10] Mono is just bad at making apps for windows work. [17:39:20] Apps writen on linux work pretty well. [17:39:22] hm, on windows I always use .net, never mono [17:39:25] OrenBochman: back if you had more questions... i haven't quite made it through scrollback yet [17:39:31] .net is better than mono on win [17:39:58] it's like using gcc instead of msvc [17:40:16] Tbh I never develop for windows... there are like 2 people in the company that use windows on their desktops and everything else down to the phones are linux or unix based. [17:40:29] but like 98% of wikipedians use it [17:40:53] so developing for a wikipedia usualy needs some win knowledge [17:40:59] I mean wikipedia users [17:41:06] huggle etc [17:41:15] if it wouldn't work on windows no one would use it [17:41:22] is HUggle still in VB? [17:41:24] no [17:41:25] c# [17:41:33] It's a weird choice but I suppose easier than say py2exe'ing stuff [17:41:46] isn't that a python huh? [17:41:54] I don't want to use it :D [17:41:56] * Damianz shrugs and thinks this is probably why he doesn't have svn access or anything pass rollback/reviewer rights on wikipedia. [17:42:05] :D [17:42:12] you don't have svn because you didn't apply for it [17:42:27] Indeed :P [17:42:36] I didn't apply for anything else either, I just know some nice admins. [17:42:42] heh [17:43:18] On the subject of svn... hmm where did I put that diff of not having a puppet clue. [17:43:28] Damianz, trust me - collecting hats is easy [17:43:45] harder is to keep basic sanity under all this load [17:43:47] MaxSem: I have too many hats, they clutter my free time :P [17:43:54] Sanity? What is this [17:44:54] * MaxSem recalls his last couple of months on Wikipedia... that was terrible [17:45:22] why [17:45:41] luckily, I resigned before becoming the first rogue скфян ыеуцфкв [17:45:53] petan: Does http://sprunge.us/ZIgT look about right for adding a edit time check to cb? (pretent it's a nice coloured diff, cause I don't have git access :P) [17:45:59] *rogue crazy steward [17:46:13] MaxSem: I think all the admins are on a certain level of crazy [17:46:17] :D [17:46:22] omg [17:46:27] lol [17:46:36] that's a diff... [17:46:41] I had too many ways to exhaust my mental resources [17:46:45] How else do you want it? :P [17:46:50] :) [17:46:54] I'll send you it as byte compiled code if you want [17:46:56] >.> [17:47:00] in human readable form :D [17:47:16] nvm it's ok [17:47:30] Damianz: want to commit it? [17:47:58] /^-/d | s/^+// ? :P [17:48:28] It is committed? I'd push it for review but as noted previosly gerrit hates me. I'm not 2% sure it's right as puppet has a weird syntax. [17:48:40] ah [17:48:57] push it [17:49:04] then I will read it :) [17:49:05] I was going to add it as a misc class but apparently adding a new host class is the way to go =/ [17:49:05] heh [17:49:50] I don't think that php packages should be defined like this [17:49:58] rather insert them to generic packs [17:50:10] dunno [17:50:39] to generic-definitions.pp [17:50:42] * Damianz wtf's at git [17:50:47] heh [17:50:50] !git [17:50:50] for more information about git on labs see https://labsconsole.wikimedia.org/wiki/Git [17:51:11] >.> [17:51:11] btw this bot is running on mono too ^^ [17:51:17] wm-bot [17:51:30] that's why no noe wants to have it in #mediawiki heh [17:51:33] I was thinking more about it trying to get my key pass from gnome-ssh-askpass when I have a ss-agent running... [17:51:43] ah [17:51:54] you want too much [17:52:33] btw am I only person using kernel 3.0.4 on debian? :D [17:52:44] all folks around have 2.6x [17:52:58] I seriously had troubles with that [17:53:08] this new one runs so much better with my hw [17:53:45] I usually run the rc releases or just head. [17:53:55] I run anything that works [17:54:03] but I run a lot virtual stuff [17:54:18] and older kernels have had terrible support for that [17:54:29] meh [17:54:50] back to git [17:58:32] New patchset: DamianZaremba; "Adding generic check for user's last edit time + adding 2 checks for the bots-cb server along with some package requirements. The bots-cb manifest will be expanded upon once the core components are packaged up and we can properly roll/configure a cb serve" [operations/puppet] (test) - https://gerrit.wikimedia.org/r/1712 [17:58:37] Ah [17:58:38] ok [17:58:47] Why is the git username the postix one but the web one the wiki one [17:58:53] * Damianz grrs and runs around [17:59:07] heh [17:59:12] I think you switched them? [17:59:18] I don't know [17:59:19] nope? [17:59:26] gerrit uses posix I think [17:59:37] I'm logged into the web interface with DamianZaremba but can push only with damian [17:59:40] * Damianz shrugs [17:59:42] Oh well [18:00:05] * Damianz thinks he should have just got them both set to the same but was trying to keep a mw standard. [18:01:41] I'll just go finish changing the groups around on my radius cluster then come back to working out puppet. [18:02:01] hm... [18:07:10] stupid question: Perl is dual Artistic/GPL [18:07:20] what GPL is it? [18:09:08] !wiki GPL [18:09:08] http://en.wikipedia.org/wiki/GPL [18:09:51] no idea :D [18:09:56] wikipedia doesn't know that [18:10:02] i mean which version [18:10:08] no clue [18:10:15] it's not on wikipedia either [18:11:40] that's what I hate GPL for: in addition to fighting copyright it creates its own bureaucratic layer [18:12:01] and you can't even combine different versions of it [18:12:23] so my extensions are WTFPL [18:12:46] I am using gpl v3 [18:15:16] MaxSem, you can combine if they have the "or later" [18:16:05] =if someone didn't forget to put that "or later" in the proper place [18:24:38] GPL is weird [18:24:40] WTFPL ftw [18:24:56] Though the real question is, is it free as in freedom or free as in beer... or free as in both. [18:25:06] !wiki WTFPL [18:25:06] http://en.wikipedia.org/wiki/WTFPL [19:03:54] PROBLEM Disk Space is now: WARNING on nova-production1 nova-production1 output: DISK WARNING - free space: / 565 MB (5% inode=86%): [21:00:46] I hope my carpet likes mtn dew =/ [21:58:54] PROBLEM Disk Space is now: CRITICAL on nova-production1 nova-production1 output: DISK CRITICAL - free space: / 281 MB (2% inode=86%): [22:06:24] * Damianz pats labs-home-wm and tells it not to eat all the cookies [22:06:30] Hi Ryan [22:06:33] hi there [22:08:45] * Damianz humms and waits for g++ to hurry up [22:39:27] New review: DamianZaremba; "(no comment)" [operations/puppet] (test) C: 1; - https://gerrit.wikimedia.org/r/1711 [22:40:25] Hmm is there any nice way to syntax check the puppet file before I push it and gerrit moans? [22:40:33] * Damianz thinks he'll fix this later