On Fri, Sep 5, 2014 at 1:48 PM, John Mark Vandenberg jayvdb@gmail.com wrote:
IMO the WMF should stop focusing on English Wikipedia as a target deploy site, and stop allowing its product management team and WMF staff in general to be salesman for it - it is scaring the community that all WMF staff seem to be so heavily vested in this 'product' as the salvation of the wikis.
This is rank hyperbole.
The MediaWiki deployment train delivers new software to all projects every week. One stage is to non-Wikipedia projects, which actually get new software *first.* Then in a second stage is for all Wikipedias simultaneously. So the default behavior for rollouts, if all you do is merge your code and wait, is that English Wikipedia gets basically no special treatment..[1]
Now, for larger feature rollouts like VisualEditor or MediaViewer, the testing stage and eventual launch set their own special schedule. We have used English Wikipedia as a testing ground a lot in the past, which is natural when you consider a variety of factors.[2] That doesn't mean we haven't worked hard to test things out with non-English projects. Some examples:
-- MediaViewer spent a long time being tested outside English Wikipedia before it ever touched that project. It started with pilots in non-English Wikipedias and English Wikivoyage.[3] -- Flow is currently soliciting editors on non-English projects to test it out voluntarily on a sub-set of pages. Any projects that want to help shape the future of this software should pick a discussion space they want to use for testing and speak up. -- My team (Growth) has begun waiting for translations of experimental interfaces, so we can A/B test in many languages simultaneously. We're about to do this again in this month, by testing task recommendations in 12 languages right from the start.[4] We've done with other projects as well, like A/B testing changes aimed at anonymous editors in four languages. -- The Content Translation project is starting with Spanish and Catalan Wikipedias.[5]
After we get past the testing stage, none of this erases the fact that English Wikipedia is still the largest project by far, and is a major problem spot to be dealt with regarding issues related to new editor acquisition and retention. The data clearly suggests that it's a project we should be focusing on if we want to fix these problems, but we're certainly not ignoring others.
1). wikitech.wikimedia.org/view/Deployments 2). All technical staff and community contributors share English as their working language. Software gets built in English first obviously, so we don't have to wait on translations as a blocker for deployment. English Wikipedia is also our largest project, so we can get larger randomized samples during A/B tests. Making A/B tests shorter while also retaining accuracy is good. 3). https://www.mediawiki.org/wiki/Multimedia/Media_Viewer/Release_Plan 4). https://meta.wikimedia.org/wiki/Research:Task_recommendations/Experiment_one 5). https://www.mediawiki.org/wiki/Content_translation/Roadmap