I'm reminded of OCaml's `let rec` which I'm somewhat surprised to not see any mention of, neither here nor the w3c's GitHub issues (https://github.com/w3c/csswg-drafts/issues). I was originally thinking of only treating nesting from a parent to child but the child to parent relationship does complicate things (especially since it seems much more likely with CSS to need to modify a parent - assigned from a library or framework, presumably - than typical imperative or functional programming). I originally thought a decorator-type syntax where "this definition includes nested elements" would be reasonable, since any user agent or linter or whatever could be in its rights (depending on spec) to look ahead a certain distance. Obviously this wouldn't actually be too reasonable with arbitrarily distant children.
Anyway, I also initially thought I preferred the brackets since it's explicit and some of the best wording is odd to me, but I think after considering my own `let rec` idea I prefer the optional `@nest` syntax because the only odd case which is relevant - and often enough that succinct but obvious syntax is warranted - is the child->parent->* case. That's the only time I'd want a heads up while visibly parsing css.
Anyway, I also initially thought I preferred the brackets since it's explicit and some of the best wording is odd to me, but I think after considering my own `let rec` idea I prefer the optional `@nest` syntax because the only odd case which is relevant - and often enough that succinct but obvious syntax is warranted - is the child->parent->* case. That's the only time I'd want a heads up while visibly parsing css.