[02:46:36] hi kevin_brown [08:21:51] Krinkle, have you seen the recent drama about hi.wiki admin abuse? A single sysop basically prevented most editing through abuse filter. [08:22:07] haven't heard or seen it [08:22:15] We need a tool to monitor the usage of abuse filters on all wikis. [08:22:45] Krinkle, how hard would it be to make a tool which queries all wikis and makes a table with the number of filters and total hits, querying https://www.mediawiki.org/wiki/Extension:AbuseFilter#API ? [08:23:20] The easiest and shortest thing to check would be the percentage of edits hitting a filter, but I don't see a query for that. [08:23:31] There are tools for that already [08:23:58] See #cvn-commons for an example of how such a tool can be used [08:24:07] (wRealBot) [08:24:18] Krinkle, yes, but I don't mean monitoring real time, I mean ex post. [08:24:28] ? [08:24:33] I have to check all current filters on all wikis and there's no easy way. [08:24:47] Not as soon as they're changed, but at any time. [08:24:58] And with a compact view of all wikis. [08:30:22] Nemo_bis: Check #cvn-commons just join for a sec [08:31:26] Krinkle, done, and...? [08:37:17] sorry got a million tabs open [08:37:24] so you see the reports, right? [08:37:29] yes [08:37:38] sorry, commons is busy on deleting right now :P [08:37:39] but it's not what I need [08:37:43] I mean the ones from wRaelbot [08:37:54] that bot is getting its stuff from a feed [08:37:55] yes [08:38:26] the hypothetical tool you describe would subscribe to the feed that wRaelBot subscribes to as well [08:38:39] no [08:38:48] and do so for all wikis (wRaelBot is also subscribes to all wikis, and outputs commons to #cvn-commons) [08:38:58] subscribed* [08:42:27] Nemo_bis: so by usage you mean filter modifications/creations, not hits? [08:57:37] Krinkle, no, the list of existin filters [08:58:46] Nemo_bis: Can you describe what this tool would do in practice? [08:58:55] (not how it works) [09:11:04] Krinkle, create a table which lets me see which wikis are using the AF at all or have very active filters [09:12:31] Nemo_bis: Very active in number of hits or percentage? Number of hits might raise en.wiki to the top in practice regardless [09:12:40] but yeah, that's an interesting idea [09:12:42] Krinkle, percentage [09:13:00] But I don't see any way to query that piece of info [09:13:08] (the global percentage) [09:13:13] The percentage would have to account for abusefilter hits that prevented an edit, as well as abuysefilter hits that did get through (not double counting or skipping either) [09:21:15] Krinkle, yes, that would be better, but 1) let's start with simple things, 2) you can't do it if you don't have view permission for all filters (you'd need to login with a steward account). [09:21:39] right [09:21:55] I guess you could do it from the toolserver [09:21:57] Krinkle, is there any way to get the API output for any given module include the domain name? [09:22:08] using recentchanges edit=1 and logging tables as base [09:22:13] (logging/abusefilter hits) [09:22:20] hm [09:22:27] APi would need permissions as you say [09:22:35] But isn't the log restricted anyway [09:22:41] sure, parts of it [09:22:51] but you don't need to know who did logs [09:22:55] just which logs occurred and when [09:22:58] Well, this would need more calculations, though [09:23:34] get number of edits in the last 24 hours via (FROM recentchanges where rc_type=RC_EDIT or something) and then get number of abusefilter hits via abusefilter_logging [09:23:52] assuming the percentage would not be of 'all time' but only recent [09:37:20] Krinkle, I'm stupid so right now I'm just running a wget -x --cut-dirs=1 on a list of all http://ab.wikipedia.org/w/index.php?title=Special:AbuseFilter&uselang=en etc. and will grep the HTML... [09:37:48] Is that public/ [09:37:52] Then what is the problem? [09:38:30] Krinkle, the problem is just that one can't open 700 special pages every day/month to check what's going on [09:39:18] I mean, if viewing is public. Then why would you need to wget those html things? [09:39:29] Do you have a Toolserver account? [09:39:29] And it will need a lot of manual work to convert it to a usable concise form. [09:39:53] Yes, but I don't want to run long queries nor to learn MySQL today [09:42:23] OK, I recommend creating an issue on JIRA for the query-service and ask for help with the formulation of a MySQL query that calculates the percentage of edits that are abuse filter-hit. Then once you have that query you can ask someone with a Toolserver account (like me) to run that query for all wikis every day and dump it on a page in a nicely presented way. [09:42:41] I can create such tool in a few minutes, but not the query [09:43:35] Ah, ok. [09:44:07] The query to be meaningful should span several days on small wikis, isn't that too huge? [09:44:25] should be fine [09:44:36] https://wiki.toolserver.org/view/Query_service [09:45:09] The people of the query service know the mysql databases very well, they'll be able to come up with a query that can do this. Then I can take that query and run it for you and create a new presentation around it [09:45:18] ok, thank you [09:45:30] You're welcome [09:48:58] Krinkle, no need of help for the API part of the querying (if any), right? [09:49:24] I'm not sure abusefilter even has an API [09:49:45] Krinkle, it does https://www.mediawiki.org/wiki/Extension:AbuseFilter#API [09:49:57] not a complete one but something [09:50:43] hm, nothing one can't more easily extract from the DB probably [09:50:59] yeah, DB query is better in this case [09:51:06] since you need to map it with number of edits [10:11:10] Krinkle-away, https://jira.toolserver.org/browse/DBQ-178 [16:45:30] sumanah: Is that too late to apply for GSoC? [16:45:51] hi bharath [16:45:53] no it is not [16:45:55] !gsoc [16:45:55] MediaWiki participates in the Google Summer of Code mentorship program. http://www.mediawiki.org/wiki/GSoC Please read http://en.flossmanuals.net/GSoCStudentGuide/ if you're thinking of applying to GSoC. [16:46:19] bharath: you should read that introductory material and you should read Google's GSoC FAQ, which is linked to from those pages. [16:49:53] sumanah: I have talked to you before but because of my exams I am unable to concentrate on GSoC but now I am free and want to do some android projects in GSoC [16:50:51] bharath: have you read that introductory material? [16:51:02] If you had, you would know that student applications haven't even started yet. :-) [16:52:38] sumanah: yeah, but I have heard some mentors will keep some students who approached them first and working on the projects now [16:53:35] bharath: Yes. This is true. But you asked "is that too late to apply" [16:53:49] bharath: so, tell me more about why you want to do Android projects [16:54:30] bharath: if you want to work on Wikimedia mobile stuff, there's an IRC channel specifically dedicated to that [16:58:10] sumanah : It interest me more and frankly I got lot of resources for android [16:59:04] bharath: Go talk to the folks in #wikimedia-mobile and tell them about what you're interested in, and try out https://www.mediawiki.org/wiki/Mobile/PhoneGap/Tutorial [17:01:18] sumanah:Thanks a lot [17:01:33] bharath: and make sure you ask clear, accurate questions. [17:01:48] bharath: http://catb.org/esr/faqs/smart-questions.html is a bit harsh but has the right idea. [19:39:31] have there been any discussions about allowing/supporting webgl in mediawiki? [19:45:46] Emw, supporting WebGL? [19:45:58] Emw, how could MediaWiki make use of WebGL? [19:47:01] Krenair: one of the most prominent examples that comes to mind is molecular visualization of chemical structures [19:54:07] for example, a rotatable 3d structure of a protein could be placed where a lead image would usually be in the protein's wikipedia article. (e.g. a manipulable 3d structure like http://webglmol.sourceforge.jp/glmol/viewer.html in the infobox in a page like http://en.wikipedia.org/wiki/FOXP2) [20:11:21] Emw, on strategy.wikimedia.org probably [20:13:48] Nemo_bis: it looks like "webgl" doesn't appear anywhere in strategy.wikimedia.org: http://www.google.com/search?q=webgl+site%3Ahttp%3A%2F%2Fstrategy.wikimedia.org, http://strategy.wikimedia.org/w/index.php?search=webgl&title=Special%3ASearch [20:15:46] Emw, there are several proposals here: https://strategy.wikimedia.org/wiki/Category:Proposals_for_graphics-related_features [20:16:22] ah, thank you [20:19:12] given that webgl isn't supported in internet explorer (and it looks like it won't be supported in ie10), what are the chances webgl would be allowed/supported in wmf mediawiki deployments even if prototype implementations of webgl on mediawiki were available? [20:19:25] not sure it's what you want, but the basic idea is there [20:19:41] * Nemo_bis has no clue to answer such a question [20:24:27] for example, SVG has been supported by major browsers for 3+ years, but i believe SVG images are still not supported in wikipedia (any language) because of concerns with legacy browsers (IE8 and lower). WebGL is a much newer technology. Would WebGL's support in wikipedia likely share the fate of SVG? [20:25:38] Emw, I don't know much about WebGL - is it done via JavaScript? [20:26:41] Krenair: yes, to my understanding it uses JavaScript and the HTML canvas element [20:27:20] I don't think allowing people to put JavaScript into pages is a good idea, wouldn't that allow people to put cookie-stealing code and other malicious things in? [20:34:15] Krenair: i also don't know much about WebGL, but i don't the JavaScript source code would need to be editable -- my impression is that only the inputs that define the WebGL 3D structure would be need to be editable [20:37:05] for example the WebGL structure at http://webglmol.sourceforge.jp/pymol_exporter/all.html can be changed by changing the input structural data -- and the inputs are non-JavaScript [20:38:54] the inputs would need to be properly escaped to prevent XSS hacks, but i imagine there are precedents for this in current JS-enabled features on wikipedia (e.g. tables that expand/collapse with JS) [21:10:23] Anyone around now that knows CentralAuth and might be able to help me setting it up to test with database table name prefixes? [21:49:22] Reading recent changes to https://www.mediawiki.org/wiki/Roadmap [23:40:37] Krinkle: leaving #mediawiki to avoid getting drawn into further "what wiki" discussion [23:41:12] basically: in the Gerrit migration, it's a very annoying painpoint that it's hard to find the repo viewer(s) [23:41:17] Idea: make a JS bookmarklet such that you could drag a project name from https://gerrit.wikimedia.org/r/#admin,projects and it would transform it into the right gitweb URL. [23:41:38] Krinkle: is this something that is a reasonable thing to ask you or another of our JS pals to do? [23:44:47] (hi robla) [23:44:54] hi [23:45:40] sumana-h: no need to [23:45:53] sumana-h: click 'branches [23:46:08] "(gitweb)" [23:46:25] and from there is even a direct url to the .git file for anonymous or logged in checkout [23:46:36] Krinkle: Where should I click "branches"? [23:47:09] https://gerrit.wikimedia.org/r/#admin,projects > https://gerrit.wikimedia.org/r/#admin,project,mediawiki/core,info > branches > gitweb > summarty > checkout url / repo url / tree etc [23:47:19] it's one click too many imho, but doable [23:47:41] should be listed in the overview, i agree :) [23:47:52] that's more like 3 clicks too many :-/ [23:49:31] I guess I could ask all the Gerrit project owners to put it in the Gerrit project "description" [23:49:41] so that it would show up in, for example, https://gerrit.wikimedia.org/r/#admin,project,analytics/reportcard,info [23:49:56] the description isn't parsed [23:50:03] so that'd be copy-paste duty [23:50:18] 2 more clicks, a drag, 2 keystrokes and a click :P [23:50:51] yes. [23:50:56] But at least it would be THERE. [23:51:04] Visible and obvious. [23:51:32] gerrit isn't meant to be obvious [23:51:34] :P [23:52:17] If gerrit would just have not both the logic, skinning and interface hardcoded in one piece.. [23:52:56] Krinkle: Yeah. It's really very uncustomizable. Apple-esque tyranny only works if there is a real designer.... [23:53:02] the nice thing thought that due to everything being so javascript-y in gerrit (given away by everything being in a # url), it is possible to write a different UI that interacts with the real gerrit in the backend [23:53:15] Krinkle: anyway, so, you can see why I think a bookmarklet would help [23:53:34] yeah, I'll see what I can do [23:53:36] Krinkle: well, I hear from Roan and Chad that actually this is not possible because of the way JS is delivered in Gerrit [23:53:57] I hate Gerrit's UI. [23:54:00] ("this" being writing a different UI) [23:55:50] * Krinkle tries a quick thing with the gerrit API [23:55:57] (which doesn't' exist officially) [23:55:59] ("API") [23:56:00] yeah [23:56:04] biab