See also ericniebler/stl2#625.

@CaseyCarter, your view on this?

It's not a range adaptor object, nor specified to be a range adaptor object. I do consider it nevertheless to be a range adaptor: it adapts the counted range [i, n) into the iterator+sentinel range [i, i+n) when i is random access or [counted_iterator{i, n}, default_sentinel) otherwise.

I think the source of the confusion here and in ericniebler/stl2#625 is that we've applied the too-generic term "range adaptor" - by association with the terms "range adaptor object" and "range adaptor closure object" - to a subset of the things that can reasonably be considered to adapt ranges. Ideally we'd devise a more specific name for that subset.