These changes look good, but this repository was originally a proof of concept and I don’t really want to maintain it. If you’d like to take over maintenance I can give you access to https://crates.io/crates/wtf8.

A side note on Ord / PartialOrd: sorting UTF-8 or WTF-8 bytes like this gives the same results as sorting lexicographically by codepoint. This is generally what you want when dealing with Unicode, but it is not the same as sorting [u16] UTF-16 / WTF-16 code units. For that you’d want to use Wtf8::to_ill_formed_utf16 with Iterator::cmp.

Since WTF-8 is designed to deal with WTF-16 systems, obvious which of the two orderings a user of the library might want. And so I think it might be better not to implement the Ord or PartialOrd traits at all. (Instead, maybe provide two fn cmp_codepoints(&self) -> cmp::Ordering and fn cmp_u16_units(&self) -> cmp::Ordering methods.)

(WTF-16 is when "potentially ill-formed UTF-16" is too long: https://simonsapin.github.io/wtf-8/#wtf-16)

Right, I suppose there isn't much reason to actively maintain this repo when the main consumer of WTF8 already has an internal copy of it. I might just go the same route and copy the in-tree version into my own library. I don't imagine there are very many people out there with a use-case similar to mine anyway.