Re: [csswg-drafts] [mediaqueries-5] How does script control of preferences work with forced color schemes? (#11575)

> > IUC light only was to instruct the UA to not try and automagically create a dark theme (Chrome had this [as an experiment](https://developer.chrome.com/blog/auto-dark-theme/) once, but the feature (thankfully) got pulled).
> 
> Fwiw they still have that available on Chrome for Android, and that's essentially the basis for this discussion. (Without going off topic I'd be curious to hear why you think it's good it got pulled? I use it daily and it's actually pretty sweet, borderline an accessibility feature to me).

The feature is still feature flagged (ever since 2021) and not on by default. I was under the impression that it was already gone, but apparently it’s still there – Oops!

I (personally) dislike the Auto Dark Theme feature because having a dark theme is something that should be intentional and art-directed. I also have concerns about brand identity and integrity – as I brand I would not like to see my carefully designed logo and colors to suddenly be inverted. And from what I can tell after some quick tries is that it doesn’t work in all cases and can introduce low contrast situations. Try it out on `lidl.be` for example – that site that has no dark theme and Auto Dark Theme ain’t really doing it.

> My question was more around what the behaviour the override has irrespective of the meta tag aspect (for the sake of discussion lets imagine the site defines support for `light dark`). Should the override pick "light" or effectively "light only" such that the browser wont still try to force dark.

Ah, I see what you mean now.

In my mental model for this, this is setting inherits down from the OS down to the site, with the last level explicitly setting preference winning.

1. Operating System
2. User-Agent
3. User
4. Site

See https://brm.us/dark-mode-toggle#inherited-settings + a bit further down for a POC _(I only use three levels in that POC, but consider the last one the User as I am assuming that there is a 4th level in which the site itself shares which values it supports)_

So for example with the following values, the result is `dark` (because the User said so, and the Site doesn’t care):

1. OS = light
2. UA = inherit _(= I don’t care)_
3. User = dark
4. Site = light dark _(= I don’t care)_

So for example with the following values, the result is `light` (because the Site said so):

1. OS = light
2. UA = inherit _(= I don’t care)_
3. User = dark
4. Site = light only

And here it’s `light` (because the OS said so):

1. OS = light
2. UA = inherit _(= I don’t care)_
3. User = inherit _(= I don’t care)_
4. Site = light dark _(= I don’t care)_

So I don’t think there’s a need for using `only` in the overrides here?

-- 
GitHub Notification of comment by bramus
Please view or discuss this issue at https://github.com/w3c/csswg-drafts/issues/11575#issuecomment-3427207854 using your GitHub account


-- 
Sent via github-notify-ml as configured in https://github.com/w3c/github-notify-ml-config

Received on Tuesday, 21 October 2025 15:15:50 UTC