Lots of documentation examples still use encoding APIs with encoding names
rather than encoding constants. I think it would be preferable to direct
users toward constants as it can help with auto-completion, static analysis
and such.