> > Does anybody want to defend the status quo?
>
> Not me. Do we need separate __cpp_lib_constexpr_chrono (etc.)
> macros, then?

My impression is that either we need to go back to that or we need to pick a
name that we're more confident we won't want to use for some other purpose
in the future. Or, as Matt suggests, maybe we don't even need a macro for
this purpose. But going back to header-specific names seems safest to me.

Apparently we changed to what we have now after Bristol. Here is the only
relevant comment on the Bristol wiki:

Benjamin: Maybe only have one macro __cpp_lib_constexpr, not individual
macros for each chapter. Do not go sub-modular right now.

All things considered, I think we should more or less reverse that direction.

But the actual names and header associations on the Bristol wiki were a
feeble effort. Here's what I would suggest now:

N3471 __cpp_lib_utility_constexpr_functions <utility>
N3469 __cpp_lib_chrono_constexpr_functions <chrono>
N3470 __cpp_lib_array_constexpr_functions <array>

Would anyone (especially Benjamin) have an objection to this direction?

Clark