We don't want to index exposition-only names at all, I think, because those are not names that could appear in user programs.

So, exposition-only member functions should not be indexed.

For exposition-only classes, we do want to index them, if/when they contain user-visible member functions (e.g. operators in case of views).

@jensmaurer What about exposition-only members such as move_only_function::is-callable-from?

And namespace-scope variable templates such as is-vector-bool-reference?

All kebab-case names should be in exposition-italics. If you notice any that aren't, we should fix them. But those exposition-only names cannot ever be called by users, so they shouldn't appear in the index of library names.

@tkoeppe, we need a new issue here that we can mark "decision-required". I do agree the exposition-only names should be in some index (or several?); the question is which one. I'm not strongly opposed to the library names index if we add some intro paragraph explaining the typeface.

We can open such an issue, but for now we have four straggling exposition-only names that remain in the library index, in the wrong place (namely right at the start), which doesn't seem all that useful, either.

Maybe we could put all the exposition-only names into the existing library index, but into a separate section?

We should sort them in the proper alphabetical order; anything else amounts to a special index that nobody wants.

The library index also serves as a list of reserved names (e.g. for the purpose of defining macros) so maybe there's some value in not "polluting" that list with names that don't affect actual programs, but only exist for the purpose of specification?