Universal selector is described incorrectly

[annevk]

https://drafts.csswg.org/selectors/#the-universal-selector says it's written like a CSS qualified name but that is wrong, as * is not a valid CSS qualified name. \* is, but is not the same as the universal selector.

Gecko appears to do this correctly for both the prefix and local name. Chromium does it correctly for the local name, but not the prefix. WebKit currently treats * and \* as equivalent, for both local name and prefix.

There's no test coverage for this currently.

This relates to #8909.

(https://drafts.csswg.org/css-namespaces/ seems correct in how it describes this, though there might be room for improvement.)

(I discovered this mismatch in https://bugs.webkit.org/show_bug.cgi?id=184604.)

[annevk]

I tend to think they should be the same. For getElementsByTagName() and XPath there exists no syntactic distinction between * meaning all and * meaning an element named *. And even when we have more liberal element names as per whatwg/dom#849 (comment) you would not be able to name an element *. So on balance I think it makes sense if * and \* end up parsed in the same way. And it would allow implementations to reuse the qualified name representation they already need.

[dbaron]

web-platform-tests/wpt#40451 adds a test with a TODO comment pointing to this issue.