@Jdlrobson - do you know if it's possible set up an a/b test between the new and old styles? Can we send multiple styles to the browser?

Not really. In theory we could do this but it would mean limiting print styles to users with JavaScript and/or double loading them. Since we are targetting users with inferior browsers this might not be the best of ideas.

What would such a a/b test look like anyway? What would be the proposal? Most importantl, why do we feel the need to test this?

@Jdlrobson - since we'll be making a large change to our print styles, we'd like to see if this has any effect on how often people print, with the hypothesis that the new print styles do not cause users to print less often (i.e. we haven't accidentally removed something that was valuable)

I don't think it's worth trying to measure anything else. If we can bucket users, the a/b test would look something like (@Tbayer - feel free to add any corrections to the outline below):

1. Deploy new styles to all wikis
2. Assign 50% of users to the new styles and 50% of users to the old styles
3. Track the rates of usage for each group (how often they print)

If we can't do this with the browser print, perhaps we can also set up something similar to the OCG/Electron rendering test on mediawiki from the printable version link (although we should first double-check if anyone is using this link at all) - randomly assign the old or new styles when the link is clicked and track that. Perhaps we can just set up a grafana board similar to https://grafana.wikimedia.org/dashboard/db/mediawiki-electronpdfservice?orgId=1&from=1484922362966&to=1495290362970

We have no a/b test infrastructure to ship different styles.

Option 1: we can a/b test logged in users (no caching for those users).
2: We can make styles JavaScript only but this would break the experience for users without JavaScript

2 seems irresponsible to me; will probably not go down well with members of community and i wouldnt recommend.

I should remind you that setting up analysis is expensive and may delay the project.

@Nirzar, here is a list of articles for which it would be good to get design input on: T167117. This will help us to know more about print styles before starting work on them.

@Nirzar and @Jdlrobson

We should add https://en.wikipedia.org/w/index.php?title=MediaWiki:Print.css&diff=805936765&oldid=805591761 this to the stylesheets. Otherwise pages with lots of floating content like https://en.wikipedia.org/wiki/Template:Didcot,_Newbury_and_Southampton_Railway, have that black bar and licensing info next to the floating element. which isn't particularly nice usually.

@TheDJ sounds good to me. If you submit a patch I can happily merge. Otherwise, hopefully I can remember in 2 weeks time (team is about to pause for an off site)

@Jdlrobson What about the child tasks in reach/out of reach?

• T169828
• T171114
• T171330
• T171023

Creating a new epic should be done now.

I don't think an epic is useful but I've pulled out the remaining tasks into a tracking task to make them easier to find > T241376