> Is the timezone argument to make-timestamp a timezone object, or an
Olson string or an implementation-dependent string?

It is a timezone object. However, the tz-timezone procedure takes an
Olson string and returns a timezone, but I think I might loosen it to
allow for implementation dependent strings.

The reason is, is that the zoneinfo directory of many systems has names
that are not in the Olson style. Examples include "EST5EDT" and "ROK"
for backwards compatibility. And some operations might use different
identifiers, as you said.

-- Peter McGoron