This is probably the spec reference we'd consider authoritative: https://drafts.csswg.org/css2/generate.html#propdef-content

One option is obviously just getting this added to CSS 2.2.  Greg, any preference one way or the other?

I'll leave this Untriaged for the CSS experts to look into.

Oh also, do you have examples of sites hitting this?  I see you marked it Pri=3 so guessing it's not a common source of interop issues, right?  If we've seen it cause problems more than once in the wild, I'd suggest upgrading to Pri-2.

Context: https://twitter.com/gregwhitworth/status/712800048837341184, including:

Mike Taylor ‏@miketaylr: that quirk prevents us from shipping `-webkit-device-pixel-ratio` cf., https://bugzilla.mozilla.org/show_bug.cgi?id=1237101#c6

Google docs is apparently depending on this quirk somewhere - we should either standardize the API or break it.

Sounds like may be Pri-1 (or at least Pri-2) to me.

Also WebKit issue: https://bugs.webkit.org/show_bug.cgi?id=155820

FWIW, Presto-based Opera 12 used to support generated content on any element, including text content (not just `url(foo)` like in Blink).

E.g.

    html {
      content: 'Yay, CSS!';
    }

CSSWG has always intended to have 'content' work on every element, so I'm in favor of us keeping this behavior, and getting the specs to align.

That said, our current behavior violates what the spec would say - as Comment 8 says, we only allow images when applied to arbitrary elements, when we *should* be allowing text as well.  

Also, for both pseudo-elements and normal elements, we have the magic "if the value is a single url(), treat the element as a replaced element" behavior which doesn't even match the 2.1 spec.  That said, this is clearly a good behavior that we should get the specs to support; users that wanted the old behavior (of the image being a replaced-element *child* of the container) could get it with "content: url(...) '';" - just use an empty string to opt into the generic "this is all child contents" behavior.

Actually it looks like there's an updated spec for this here: https://drafts.csswg.org/css-content/#content-property

It says:
"content applies to: ::before, ::after, ::marker, and page margin boxes. Image and url values can apply to all elements."

So this is WAI according to the current spec, right?  Tab / Greg?

Yeah, at the last CSSWG f2f, we ended up standardizing on the WebKit/Blink behavior.

Closing as Fixed as per Comments #11, #12. Please change the status if I am incorrect.

Perhaps a WontFix is more appropriate here, as nothing was changed (no code changes).

Yup you're right. Thanks!
Changing the status.