[css-pseudo-4] "Singular" vs "plural" pseudo-elements

[tabatkins]

We discussed in the F2F that a few things being worked on need to be able to distinguish whether a pseudo-element is "singular" or "plural" - that is, if you target the pseudo-element (via a selector, a getComputedStyle() argument, etc) from a particular originating element, will you necessarily get 0/1 pseudo-element, or is it possible get N>1 pseudo-elements?

For example:

• ::before is singular - only one exists on an element.
• ::view-transition-group() is singular if a VT name is used (because names have to be document-unique), but plural otherwise (multiple VTs can have the same VT class, or the * argument matches all VTs).
• ::part() is always plural - the part name has the semantics of a class, and can refer to any number of elements.

We should reify this concept in Pseudo, and specify that APIs like getComputedStyle() can only refer to singular pseudos, or conditionally-singular pseudos if certain conditions are met.

[schenney-chromium]

I would phrase it as "... supported when the style can be uniquely defined, either because there is only one (such as ::before) or because you must have a name that identifies the pseudo (such as ::highlight). Cases where there may be multiple do not get reported in getComputedStyle, even if there is only one that applies to the element."

That is basically a "yes/no" to whether support exists, with a way of explaining what the answer should be.

The alternative in my mind would be returning a list, but that's overkill to me.