It's business practices seem distasteful, several legal threats against competitors and critics, heavy covert "submarine" advertising (i.e. ads masquerading as news), and apparently unexplained popularity among policy-makers despite criticism from those on the front line (e.g. emergency services). Technical criticism centres on the fact that it's algorithm is not as good as it could be, hence risking lives, and is not open. Several alternatives exist, some of which the creators claim address these technical issues:
What3Words is an interesting idea, but the lack of locality and propriety nature of the codes makes me doubtful of using it for anything important.
I prefer the pluscodes introduced by Google as it is open source, useful offline, conveys locality between codes, and doesn’t hinge on the use of English.
I made something similar that has locality https://wherewords.id but that sometimes gets criticised. Some people believe that not having close positions be similar makes it less likely to make mistakes. Personally I'm skeptical about this, but it's certainly an argument that gets made.
I’ve heard that argument. I’m also skeptical about its weight compared to the benefits of locality. When what3words says it, I feel like it’s a salespitch as to why we should use their proprietary solution.
I think your checkmoji solves that problem in its entirety more elegantly. Would a fifth word make it error correcting rather than error detecting?
I did look at error correction - it'd be cool to have e.g. a any 4 from 5 words scheme, but for all the schemes I know, you'd need a good few more words than that to make it work.
I actually have come round really to thinking that a fifth word is probably a better approach than the emoji, although the emoji is more fun. Problem with emojis is that not all of them are easy to describe, and I think audio/speech rather visual is the key here.
Each word for my word list is 12 bits of information, so for a check word, it could have 3 bits for error checking each of the location words. which would certainly let me highlight which word was wrong, and may even enable me to suggest words that might correct it.
So I'd been thinking about something fancy like Shamir's secret sharing or some other threshold scheme, but actually, an xor check word is still really good - not only does it check, but if you know that only a single word was wrong and which it was, it can recover that word, or if you don't know which word but are confident it was only one, then you can narrow it down to just 4 possible locations, which will probably be good enough to identify the actual location most of the time.
What3Words has been discussed dozens of times on HN, but the most relevant one is probably this analysis of the algorithm and the properties of the generated codes, as well as its companion piece:
Personally I think the *idea* is great and is actually useful but the closed and internet only nature of this service is troubling and the behaviour of the company doesn't seem that positive. There are better solutions which run with the same idea. This company is VC backed and is using (it seems to me) all of their funding for marketing.
The better solutions don't have vc backing and no millions to spend on marketing.
I have an old and unfinished article [1] that uses geohash / quadtiles to implement the same thing.
If your goal isn't "every location on Earth", you have a smaller bounding box (like a country or "all of Europe") and this will result in the squares being more precise, or the identifiers or word lists being shorter. Could be worth the trade-off depending on the requirements.
When I made a location to word system I ended up having to spend a huge amount of time on the word list to try to remove words that tend to combine negatively. Partly because of that (but also because I went slightly more accurate than w3w and with a hierarchical encoding) I needed 4 words per location.
https://news.ycombinator.com/item?id=37359256
https://news.ycombinator.com/item?id=27350975
https://news.ycombinator.com/item?id=27058271
It's business practices seem distasteful, several legal threats against competitors and critics, heavy covert "submarine" advertising (i.e. ads masquerading as news), and apparently unexplained popularity among policy-makers despite criticism from those on the front line (e.g. emergency services). Technical criticism centres on the fact that it's algorithm is not as good as it could be, hence risking lives, and is not open. Several alternatives exist, some of which the creators claim address these technical issues:
https://news.ycombinator.com/item?id=20970664
https://news.ycombinator.com/item?id=31829267
I've no horse in this race, but frankly what3words seems a bad choice and an unpleasant organisation.
Edit to add links to reported issues and legal actions:
https://archive.is/6KVvL
https://techcrunch.com/2021/04/30/what3words-legal-threat-wh...
http://web.archive.org/web/20210430005537/https://twitter.co...