Indeed I do! Thanks for the link. It's fun to see how differently the implementations evolve when the goal is to optimise for footprint over speed. I guess I've supplied the missing runtime-optimised versions, just a few years late. :)

(I also observe in passing that the implementations in that thread are all buggy in the general case. They don't take into account that "foo bar" and "foo bar\x00" will pack to identical strings. Not a problem if you know there are no nulls in the input, but that's why the length byte in my data gives the number of packed characters rather than simply the length of the field.)