[css-color-adjust] Re-add `only` to mean "don't auto-adjust me", per WebKit's behavior

[tabatkins]

Last year, in issue #3881, there was some side discussion that suggested WebKit was doing reacting to 'color-scheme' in a way not described by the spec. I summarized it in this comment:

We described this last year in #3299, which is still open. Indeed, it does look like that original intent (color scheme transformations at least) was lost in translation.

Ah, you're referring to "If the user's preference does not match something in the list, the UA is allowed to apply transformations to the content (e.g. apply high-contrast)." from Simon's post.

If that's the extent of it, opting the page into forced-colors mode, then that's fine, it's just a separate feature that's technically orthogonal to this.

But if y'all are planning on doing something separate from forced colors, please actually describe it in a new issue so we can talk about it. We're trying our best to make sure these features are described accurately and fully, so authors can have a reasonable idea of how their pages will look. If the "same" feature ends up still acting substantially differently between browsers, then there's no point to it being standardized at all.

No one from WebKit responded after that, so I'm still not sure whether WebKit implemented 'color-scheme' as described in the spec (+ maybe an (allowed) switch into "forced colors mode", as described in the spec, if the user preference isn't reflected in 'color-scheme'), or if they're doing something novel and not reflected in any spec.

@smfr, @hober, @dino, can y'all figure this out for me? WebKit invented this feature and shipped it before a spec existed, so we've been forced to specify it after-the-fact by reverse engineering. If we've gotten it wrong, please tell us how.

[smfr]

On the web, WebKit has light and dark modes, and no forced colors. In apps like Mail, WebKit can do some additional forced-color-like things, for example applying a color filter to invert CSS colors. I believe this spec text was to keep the doors open for something like this. I believe Chrome has also been looking into "auto" dark mode, which is similar.

[xeenon]

Yes, the color-scheme property allows iOS and macOS Mail know which color scheme to use for an email message, as it can affect the auto-darkening transformations that get applied to the message’s colors. Declaring color-scheme: light dark in a rich email message lets Mail know it supports its own styling for dark mode. You can also specify light only, informing Mail that it should not transform your light color scheme email message.

[tabatkins]

Okay, adjusted the title to reflect this.

Now that we've dropped the spec's previous definition of "only", which did not do what y'all are describing, we can discuss re-adding it to do what y'all are describing. ^_^

[smfr]

I'm not sure how "only" morphed. The original proposal has:

only
Ask the UA to use one of the listed color schemes when the UA preferred color scheme is not listed in the supported color schemes. This can typically be used force a dark theme rendering when the preferred color scheme is light.

[tabatkins]

It didn't morph (from that definition). That definition is "if the user preference doesn't match author preference, the author preference wins", which is exactly what the spec recently said (before I removed it this morning). That's also exactly what Safari does, if I'm understand you correctly. (Now it's the behavior of the spec in all cases, fwiw.)

The difference is if the author and user preferences don't match, but the author didn't use the "only" keyword. Rune reverse-engineered the algo (8.2 in his proposal) to be "use the browser default in that case". Safari apparently actually does "auto-fix the author's colors to match the user preference".

This issue is now about: what are those auto-fixes, and should we specify that in the spec?

[AmeliaBR]

So, if we were to spec the Mac Mail behavior, my current understanding is that it would be something like:

‘only’
If the user's preferred color scheme is not among the listed color schemes (in the color-scheme property), the user agent may switch to forced-colors mode or otherwise modify the displayed colors to meet the user's preference (e.g., by using an inverted-color display). The only keyword indicates that this type of modification is likely to be detrimental to the user experience.

There would need to be a note warning that only is just a suggestion: user agents may still use forced-color mode if that's what the user needs, and that all web content should be designed with that in mind. There should also be examples of sensible uses of only (e.g., color swatches on an store website, partially-transparent images that would disappear if the email's background color were inverted).

One other complication that would need to be defined: if you use only on a part of the page, how does that interact with forced-color-adjust, if a forced-color like effect gets applied to the rest of the page?

[AmeliaBR]

Also…

Since forced-colors is a media query, which could be used to set color-scheme, which could affect whether or not the browser decides to opt in to a forced colors-like scheme, how do we deal with this circularity? Maybe the decision to do a fix-up / forced color change should be based on the color-scheme meta tag only, and forced-color-adjust (rather than an only keyword in the color-scheme property) would still be required if you wanted to opt-out parts of the page.

Do the Mail / WebKit color adjustments currently apply on the whole page only, or do they support the CSS color-scheme on subtrees?

[lilles]

Android WebView has an API for doing auto-darkening for the whole document. We opt out of auto-darkening if the used color-scheme on the root element is 'dark'. The 'only' keyword has no effect.

[xeenon]

Mail / WebKit color adjustments currently apply on the main frame only.