Geee, I didn’t know that even CSS 2.1 had a counters() function (I only used and knew the counter() function). Look at this:
The ‘counters()’ function generates a string composed of all of the counters with the same name that are in scope, separated by a given string.
The following style sheet numbers nested list items as “1”, “1.1”, “1.1.1”, etc.
OL { counter-reset: item }
LI { display: block }
LI:before { content: counters(item, ".") " "; counter-increment: item }
However, concerning
It seems that W3C (or at least one of their editors) thinks (1) could be a legitimate marker.
I doubt that the W3C or their editors implies or wants to imply any assertion about “typographical legitimacy” with the examples they include in specification (drafts). In particular, the example above is probably not meant to indicate that using “1”, “1.1”, “1.1.1”, etc. was generally a recommended or even sensible way of labeling nested list items (though it actually could be, eg in a TOC represented in HTML as a <OL>, for example).
@counter-style is in the CSS Counter Styles Level 3 Candidate Recommendation. Nowadays (say, since 2010) a W3C CR, especially if released by the CSS WG, is much more mature and well-tested than most RECs in the past.
There is a problem referring to those list items (from outside of the list) if they’re considered purely presentational. The use of type attribute implies that the list type has meaning beyond further styling - even if that styling overwrites the initial presentation (but the designer probably shouldn’t override it if it is referred to). So, with regard to just the list type (not styling) I’d say we’re still in scope for the CommonMark specification. I want to be able to talk about list item (b) in another part of the document and simultaneously create that that list item using b).