[scoping] Consider making `:slotted(...)` a pseudo class

[devknoll]

I'm having trouble finding prior discussion and I'm a noob and not a spec person, so apologies in advance.

::slotted(...) selects real elements within the same shadow tree as the selector, so it's unclear to me why it should be a pseudo element while e.g. :has(...)/:where(...)/:is(...)/:host(...)/:host-context(...) are all pseudo classes. One real world issue with this is that I can't querySelectorAll inside a shadow DOM for a ::slotted() selector, despite the selector targeting real elements by definition.

In contrast, ::part(...) being a pseudo element makes sense, since the element is essentially abstract from the point of view of the selector, since the actual element belongs to a different (shadow) tree.

There's probably some underlying reason here that I'm missing (and I would very much appreciate learning something new!) but otherwise this seems like it could be a nice low hanging improvement.

[devknoll]

If I had to guess, it was done this way because technically you are selecting a slot element with the ::slotted pseudo element. The slot belongs to the tree with the style, but the slotted content doesn't. I guess this makes sense.

Still sort of confusing though 😄

[devknoll]

Some prior discussion and rationale here: https://github.com/whatwg/dom/issues/463