> It's worth reminding any readers that what3words is a horrible coordinate system for human usage due to its unreliability. The key flaw is the choice of words that are easily confused by humans such as including both the singular and plural of the same word (e.g. likely.stage.sock and likely.stages.sock are on the opposite sides of a river) which are easily mixed up both when spoken/heard and when remembered/repeated, and also many words which sound similar e.g. 'innocence' and 'innocents', 'wants' and 'once', etc.
This is already an issue that is easily resolvable: Instead of 3 words, use 4 or 5 words from BIP 39's far less ambiguous list of 2048 words.
https://github.com/bitcoin/bips/blob/master/bip-0039/english...
Based on what3word's claim of 75 trillion squares, 5 words are needed to cover all of those squares:
https://en.wikipedia.org/wiki/What3words#:~:text=a%20grid%20...
log(75_000_000_000_000) / log(2048) ~= 4.19 < 5
If i18n is a requirement, the development of translation tables would be needed. BIP39 already has that for 8 other languages, but there's still a lot of other languages that could be added in.
https://github.com/bitcoin/bips/tree/master/bip-0039
I'm guessing that it can't be resolved due to legacy reasons, but that's just all the more reason for an open service to take over what3words' job. There's also the debate of length-vs-ambiguity, but the scenarios where the addition of 1 or 2 words will make a major difference are few & far between.