I have at least one question based on the discussions related to HTML in phabricator: What is the process by which we decide to change MediaWiki's compatibility? Is it well-defined? Does it need calling out on this page?
Topic on Talk:Compatibility/Flow
The process is Requests for comment, as most technical decisions.
Typical stakeholders are Product, Analytics, and SRE.
Typical questions/criteria are: monthly page views, edits, ability to upgrade (e.g. which OS is typically associated with that browser, could the user upgrade?), cost/benefit in terms of engineering.
At the end of the day though, the reasons can vary greatly so there thusfar hasn't been a fixed process. Sometimes it's motivated by security, sometimes by performance, sometimes by cost of development or testing, sometimes by product needs for newer abilities where it is difficult or expensive to have a fallback, etc.
If you're mainly looking for how to make sure you are aware of proposed changes, then watching TechCom-RFC on Phabricator, or looking for TechCom digests on Wikitech-l will suffice.
Indeed, I do watch the Techcom project on Phab. Most interested in having this page say why and when such will change. A sentence ad hoc gist might be reasonable if that's the answer right now.
Incidentally, one might ask whether it makes sense to have 2-4 separate RFCs all related to basic technology assessment, as are present now :^). Calls into question the efficacy of ad hoc assessment, rather than some other possible rules, such as:
- A hard rule like "modern browsers are releases from the past 6 months, basic browsers the past 5 years, unknown everyone else", or,
- A standard process rule like "we assess target browsers once after each LTS release and then stick to those target browsers until the next LTS release".
There's a similar inefficiency in PHP version planning apparently, what with the last-minute decision on which PHP version the 1.35 LTS would require (which should really have been sorted as soon as 1.34 was cut for release). (At least there was intent to plan there...).
Thanks for responding!
We did this for PHP indeed, through Support policy for PHP, which documents what we want to support at minimum and for how long. This makes it easy for administrators and users of the software to set expectations and plan ahead, while also allowing uncontroversial removing of support for versions that fall outside those criteria. A larger conversation is only be needed when considering to support something longer than the policy asks for, or to e.g. remove support ahead of time for some reason.
For browsers, I don't think this is as easy. For PHP, the developers using MediaWiki can choose to postpone an upgrade until they are ready to upgrade. For browsers, users have no control over the MediaWiki version that is deployed. They cannot keep using the old version that supported their browser until they are able to purchase a new phone, for example.
As such, I don't think critiria based on version numbers or age should have a strong influence. But, I do agree that it would help if we had something written down about what our general intention and aim is for browser support. For example, aiming for X per cent of audience on Modern, and aiming for < Y per cent prior considering removal of Basic support.
I will bring this up to product managers next quarter as part of T178356. One of the problems we have right now that make this difficult, is that we have a fairly large portion of browsers categorised as "Other" in our browsers dashboard, more than 9% in total are browsers that individually have <0.1% usage. I suspect this is not true, and that perhaps some bots or spiders are caught in here, or that e.g. some browser is seen as N different ones with <0.1% each instead of one larger family. I am looking into that right now. This is relevant because otherwise we could at most aim for 90% on Modern, which seems rather low.
Right, those were bikeshedding specific examples to the ideas of technical and process rules, of course, without intent on my part to suggest that those should even look like the rules. ;)
For browsers, users have no control over the MediaWiki version that is deployed. They cannot keep using the old version that supported their browser until they are able to purchase a new phone, for example.
Sure. I think the current "we guarantee JS+CSS support to these, CSS support to those, and you get what you get otherwise" is a pretty safe policy that incorporates considerations like that, to some extent. (Even on phones [Apple excepted-I saw that's 20% of mobile ~= 13% total], users increasingly have access to evergreen rendering/script engines even if that's not the browser that came with their phone.)
Anyway, I feel sufficiently heard now. ;)