@CaseyCarter Could you have a look please?

@jwakely: is this editorial?

Yes, it seems editorial, but completely unnecessary.

There are only a few uses of as-lvalue and none of them are in noexcept functions, and none of them depend on whether that cast can throw, and users can't observe whether or not it's noexcept. I see no reason to clutter an exposition-only function with keywords that have no effect.

Adding noexcept to as-const-pointer in [range.prim.cdata] seems like it might be more useful, since we use that directly in an "expression equivalent" context that users could actually test with the noexcept operator. As currently specified you could argue that ranges::cdata is forbidden from being noexcept because it has to be equivalent to a call to that exposition-only function which is not noexcept. But implementations are always free to strengthen noexcept anyway, so I assume the same holds for exposition-only functions.