Some of the MediaWiki:Mobile.css files on certain wikis however are quite large (e.g. Most wikis appear to be less than 1000 word length (Roan checked the db). Generally the MediaWiki:Mobile.css is small but historically it can be large.Long term plan is that TemplateStyles will make this wikipage redundant.Mobile.css is small right now so on short term we wouldn't expect a big impact, but it would be prudent to run a webpagetest job on a set of wikis to evaluate the short term impact of rolling this out rather than be surprised. Current mobile CSS is 8.86kb gzipped for the Barack Obama article.it seems the additional HTTP request might not be happening any more so that's good.Through Tech-News we can announce its removal, and inform users they should request CSS changes for mobile via Gerrit/Phabricator in MobileFrontend/hacks.less or, where possible, through TemplateStyles. Leaving it as-is was a waste of effort for the communities whom took the time to migrate to Mobile.css from Common.css, which we also disabled. On the other hand, I do not oppose continuing the direction of maintaining hacks.less if you prefer that, but then we should be consistent and remove Mobile.css support from MobileFrontend. I'm sure we can find a way to make the extra request work and not regress 2s. The current situation is broken, so current performance isn't really what we should compare. It worked well for the MVP, but I do not believe this would realistically scale in the long-term to all of Wikimedia.Ĭollaboration is key, and it'll enable a long-term result that is a maintainable, user-friendly, and (eventually) performant. I don't like isolation/duplication as a way to avoid solving technical debt. From a developer perspective, the extra request may stand out visually when looking at the HTML, but I would estimate it as low-impact for the critical path.īut even with overhead. If a previous attempt caused a 2s delay on 3G, Performance Team can help evaluate alternative ways, because that should not be the case. I agree the extra request is something to be aware of, but should be minimal in terms of overhead. Provided we find that the first paint change with this change is insignificant I'm happy to go ahead with it, but I worry about the long term impacts - I'm constantly finding myself removing unnecessary css rules which either have no impact on mobile or are unnecessary and that have seemingly been copy/pasted from elsewhere e.g. To build on that, the plan with TemplateStyles on mobile was to move these styles into skins/templates where they made sense and remove the need for it altogether.
#Mediawiki common.css skin#
I'm happy to shift more rules over there or explore ways to group the stylesheet in the existing skin styles url if that's an option. The strategy so far has been to put layout styles in a hacks module. This may be less now but I still think this is unacceptable. When we last experimented with top loading this stylesheet first paint was delayed by 2s (I think on 2g) but I remember there was also a delay for 3g too due to the additional stylesheet url.