stripping all HTML from the content

@alexhollender that is not the case! we are leaving the HTML as is, and bending it to our will with CSS 🏋️‍♀️
I'll paste a link to the "read me" + fade shortly.

After poking around the summery endpoint, I realized that if we were to fix this issue there, I'm not sure what the correct response should be. The endpoint proxies the value from pageImages api, which, as seen here: https://en.wikipedia.org/w/api.php?action=query&format=json&prop=pageimages&titles=Completing_the_square&redirects=1&converttitles=1&piprop=thumbnail%7Cname%7Coriginal
returns the same non-image for "original".

from @ABorbaWMF 's screenshots, it looks like mobile and desktop use different fonts. Is that intentional?

late loading of icon and 'read me' - I know the usual way of placing an icon is generating it via JS with mw-icon, but man... can't it just be a css background image? what would the downsides be to that?

To my delight, I think this approach -- i.e. adding aria-describedby dynamically -- actually works, at least for VoiceOver on a Mac.
https://drive.google.com/file/d/1dEwRbzQfGtDaDos15s8MngJTROsGr-ZR/view?usp=sharing

I've created the branch named aria-testing for this feature.

My first impression when reading this task was that section collapsing could be achieved with CSS, avoid this issue altogether (using the :target selector for example). However, it looks like that route has been thoroughly investigated T148591 T145106 T147338 and not pursued. ☹️

@ovasileva we can only do so much... that space is there because the editors are using a {{Clear}} template that intentionally pushes down the content. 🤷‍♂️

After some discussion it was agreed that loading as much code upfront for the test would reduce FOUC, but that placing the code in Minerva would give us the most flexibility as to when the code is executed.

This feature feels like it should live in the WikimediaEvents repo to me:

Another option that was discussed was to limit the "uncleared" thumbnails to the first 2 consecutive thumbnails, so that they don't form too much of a horizontal line, like is this scenario.

This was resolved with this patch & deployed May 7, 2018
https://gerrit.wikimedia.org/r/#/c/431109/

Whatever this was, looks like it's fixed now.
Unable to reproduce...

So the way this is currently implemented, is that the search input element has a dir=auto attribute which, according to MDN, should switch the text direction based on the characters someone types in. So, if someone starts typing in Hebrew, the input should switch to rtl.

@Jdlrobson I don't think it's new, introduced here https://github.com/wikimedia/mediawiki-skins-MinervaNeue/commit/5d46c576bb4a815015bf04a6a589ccee1270fdcb
(probably not noticed since not many people use Minerva on IE11)

@ABorbaWMF thanks for shedding light on that IE11 issue, I've opened a ticket for that over here T193881