Page MenuHomePhabricator

non WP sites identifying as WP
Closed, ResolvedPublic

Description

so far from looking at the impression logs the ones I have identifying as enWikipedia are:

SimpleWikipedia (not sure best way to fix this one)
OutreachWiki
AvisoryWiki
There are also a bunch of impressions just identifying as "Wikipedia" (Not enWiki,frWiki,deWiki) My guess is those are coming from another site as well but i'm not sure which.


Version: unspecified
Severity: enhancement

Details

Reference
bz25519

Event Timeline

bzimport raised the priority of this task from to Low.Nov 21 2014, 11:18 PM
bzimport set Reference to bz25519.

overlordq wrote:

*** Bug 25520 has been marked as a duplicate of this bug. ***

since people are understandably asking what this is about I thought I'd explain it a bit more. We appear to have a couple projects that have wgNoticeProject set to Wikipedia which means that since their content language is English they are receiving fundraising banners through Central Notice when we meant them to go only to the English Wikipedia. Trying to get them to report as Wikimedia or a more appropriate project name would be preferable (and cleaner for our impression and click through stats) and Tomasz and crew were going to look into it.

SimpleWikipedia is a bit harder because it actually should be set to Wikipedia and there is no 'simple' language code for the system (so it is quite rightfully set to en)

bugs wrote:

I think the issue is because the database name for these wikis are "outreachwiki", which is the same as the Wikipedias (e.g. "enwiki", "frwiki"), so the software assumes they're Wikipedias too. In the past, we've fixed this by manually setting $wmgNoticeProject as "wikimedia" for each one that we needed to fix. See http://noc.wikimedia.org/conf/InitialiseSettings.php.txt for more information.

yea. Interestingly looking at outreachWiki that one does appear to have NoticeProject set to 'wikimedia'. I'm not sure if that was done since the last test? Because we definitely got impressions from it.

Would it be feasible to just use the dbname?

There are three different issues being discussed in this bug, so I'll respond to each.

  1. Banner impressions - Every banner request includes both the site name and the database name for the wiki. These are pulled from the wgSiteName and wgDBname javascript variables. You should be able to tell where all banner request are coming from based on the DB name value.
  1. Banner targeting - All wikis are assigned to 1 of 12 different "project" categories for banner targeting: wikipedia, wiktionary, wikiquote, wikibooks, wikinews, wikisource, wikiversity, wikimedia, commons, meta, wikispecies, and test. This is by design. We can add new project categories if needed. My guess is that wikis like OutreachWiki and AvisoryWiki should be assigned to "wikimedia" rather than "wikipedia".
  1. Simple English - Simple English Wikipedia is reporting it's wgContentLanguage as "en" rather than "simple". Thus Central Notice has no way to correctly target it. A separate bug should be filed for this.

I checked Outreach Wiki. It is reporting the following:
wgNoticeProject="wikimedia"
wgDBname="outreachwiki"
wgSiteName="Outreach Wiki"

So I don't understand how it could be identifying as English Wikipedia. Could you show me one of the impression log entries you're talking about? I wasn't able to find an AdvisoryWiki.

sorry, It isn't reporting as enWiki on the reports we're just getting all the superfluous clicks/impressions from then while we are trying to target English Wikipedia (Language= English, Project= Wikipedia) The SimpleWiki thing is an issue I know could be a problem (We don't have a language code for simple as far as I know since it's made up in the first place)

Advisory wiki is: http://advisory.wikimedia.org/wiki/Main_Page which is closed but we've still gotten clicks/impressions from it and Philippe was trying to find a way to eliminate those if we could.

Do you have any idea what project would be reporting as just 'Wikipedia' rather then enWiki,frWiki etc? We get a fair amount of those ever test as well.

Outreach Wiki does not identify as "wikipedia" and no campaigns targeted to wikipedia should be displaying there. I'll double check this during the next test targeted to en.wiki.

Simple wiki does identify as "wikipedia", as it should. However it's language code should be "simple" (which does exist as a mediawiki language code), not "en". This should be easy to fix, but it needs to be a separate bug so that we can get feedback from the proper people first, and make sure that won't break anything on Simple Wikipedia.

Advisory wiki is reporting as "wikipedia". I'll see if we can change this to "wikimedia", which should fix the problem. Let me know if you discover any more.

Outreach wiki looks good during fundraising test. Advisory wiki is fixed.

Strategy wiki looks good, Usability wiki looks good. Any others?

looking further it looks like http://quality.wikimedia.org/wiki/Portal is also showing banners (again a close wiki) removing it from CN like we did for Advisory probably best.

Removed qualitywiki from the cn cluster config. Definitely let us know if any more wikis pop up.