[css-values-5][css-namespaces] CSS @namespace rules should probably not allow URLs with request modifiers

[weinig]

The current grammar for CSS @namespace rules is:

@namespace <namespace-prefix>? [ <string> | <url> ] ;

Since it uses the <url> production, that means any URL modifiers defined, such as the request modifiers in CSS Values 5 should also parse.

I don't think it's meaningful for @namespace to have request modifiers (though I may be mistaken). I propose adding some way for specifications like @namespace to parse only the URL itself.

(cc @noamr, @annevk, @tabatkins, @fantasai)

[noamr]

It could be useful that a URL defined elsewhere (e.g. a CSS property with a <url> type) can still be used here even if it has URL modifiers, and those can be ignored.

[annevk]

I don't think so? For namespaces using <url> at all was a mistake. (It's an interesting question as to whether these URLs even get parsed and can be relative.) We certainly don't want to make them more involved.

[svgeesus]

The two URIs are treated as strings, and they are identical if and only if the strings are identical, that is, if they are the same sequence of characters
https://www.w3.org/TR/REC-xml-names/#iri-use

They are not parsed. They are case sensitive. No url un-escaping is done.

relative URIs are deprecated as namespace names

They are opaque strings which happen to be URLs

[noamr]

If it's not useful to mix and match with CSS properties etc, it's a bit weird that they're using the <url> type.
I'd probably tend to let them have their own definition of string | url() | src() or some such that is not a <url> (if we want to keep the url() backwards compat);

That definition can live in the @namespace definition or alongside the url definition, it doesn't really matter.

[annevk]

Yeah, we should probably replace <url> with <url()> and call it a day. I don't think src() is needed.

@svgeesus that's the XML definition. It seems that for CSS that applies as well and user agents cannot eagerly parse the contents of url() as a URL and have to retain the original input. https://software.hixie.ch/utilities/js/live-dom-viewer/?%3C!DOCTYPE%20html%3E%0A%3Cstyle%3E%0A%40namespace%20test%20url(%22%2F%22)%3B%0Atest%7Ctest%20%7B%20background%3Alime%20%7D%0A%3C%2Fstyle%3E%0A.%3Cscript%3E%0Atest%20%3D%20document.createElementNS(%22%2F%22%2C%20%22test%22)%3B%0Atest.textContent%20%3D%20%22test%22%3B%0Adocument.body.append(test)%3B%0A%3C%2Fscript%3E

[weinig]

Since <url()> includes the modifiers, I think you probably want this:

@namespace <namespace-prefix>? [ <string> | <url-token> | url( <string> ) ] ;

[noamr]

Since <url()> includes the modifiers, I think you probably want this:

@namespace <namespace-prefix>? [ <string> | <url-token> | url( <string> ) ] ;

I'm fine with proposing an async resolution for this. @fantasai are you OK with that as the spec editor?