CSS has very flexible selectors, so in a sense it already has some abstraction features. Those differ from what we know from other languages, but they are still very useful and should be exploited.

I was never forced to define important colors more than twice in my CSS style sheets, usually defining them just once. Maybe my document structures are too simple to get into trouble - but maybe that's the whole trick.

(i.e. producing less bloat rather than making bloat more manageable.)

I wasn't arguing that CSS does not stand on its own two feet, or that LESS is necessary. Rather, I was just objecting to the view that one should just improve one's text editing skills instead of learning to use some kind of abstraction.

Complexity is not all bloat. For a given set of data with a given list of goals, there's an optimal level of complexity. More than this and it's bloat — but less than this and you're not accomplishing what you set out to do.

There's still a maximum level of complexity for presentations (here: HTML presentation of a set of data). That level is independent of the content. Anything above that is just irritating for the viewing persons.

This is especially true for web sites which should at least look as simple as possible. So in principle, there should be always a way to produce HTML whose DOM structure isn't too complicated, no matter what the content is.