[css-nesting] Semantic of nesting selector is unclear

[upsuper]

The current spec says

When used in the selector of a nested style rule, the nesting selector represents the elements matched by the parent rule.

It is unclear whether & in a selector means an element matched by the parent rule, or any elements matched by the parent rule. This affects what would happen when multiple nesting selectors appear in a single complex selector.

For example, should a, b { & c & { } } matches nothing (in the former semantic that a & always means a single element at a time, and since an element cannot be its descendant, this selector is nonsense), or :matches(a, b) c :matches(a, b) (in the latter semantic).

This may significantly affect how nesting rules would be implemented, and I suppose the former semantic may be easier to implement efficiently.

[upsuper]

Maybe it doesn't really matter... I don't know.

[bradkemper]

It would be the :matches(a, b) c :matches(a, b) one. But it isn't nonsense. The inner a and b could match the same selector without being the same element. Imagine if it was :matches(div, span) p :matches(div, span), for instance, which is a real world equivalent. That would match div p div, span p span, div p span, and span p div. Those do actually exist in the wild.

[tabatkins]

The quoted spec text means exactly what it says - it "represents the elements matched by the parent rule", in the same way as .foo represents "the elements with a foo class". And in both examples, it makes perfect sense to repeat the selector twice in a larger selector; each instance stands for the same set, but any particular match will choose different elements from that set. (Well, most of the time; using it in function like :is() can represent the same element twice.)