[css-values-5] Behavior of ident() in at-rule preludes

[andruud]

The spec says:

The ident() function can be used as a value for any property or function argument that accepts a <custom-ident>.

Does this mean that it's not valid in other contexts? For example: @position-try ident("--p" 1) {}. If it is valid, what happens with e.g. @position-try ident("--p" sign(16px - 1em)) {}? Or @position-try ident("--p" sibling-index()) {}?

@container ident("--p" sign(16px - 1em)) {}, if valid, would need to resolve em (etc) against something that's not the container's font-size (unlike em in the query itself), since the ident is the thing that determines what to query.

[fantasai]

Seems reasonable to make this valid anywhere that accepts <custom-ident>

[andruud]

In media queries, there is no element context, but we still allow em units that resolve against the initial font-size. In general, anything you could put into an ident() (relative units, sign(), sibling-index(), etc), you could also put into a media query, and it must have (or should have) some behavior already.

So how about (proposed resoluition): ident()is allowed anywhere <custom-ident> is accepted. When used outside of an element context, the numeric parts behave as they would in a media query.

[andruud]

Ah, no. It's too much pain for zero (?) benefit. We would then need to parse ~all rules (also style rules, since some selectors can hold custom idents) into some not-fully-resolved form that needs to be (re)computed into a final rule when media values change. I don't see that actually happening.

Then I think we should just make ident() invalid without an element context. (Which may be what the spec is saying already.)