Show HN: WhatFreeWords: Open-Source Geocoding for What3Words | Hacker News Reader

Show HN: WhatFreeWords: Open-Source Geocoding for What3Words

whatfreewords.org

113 pointswhatfreewords49 comments

What3words is patented[1]. If you compare the explanation of their free implementation[2] with the claims section of the patent, it seems very possible that it breaches the patent. In particular, the equations that are described in claim number 7 of the patent are exactly the same as in their explanations.

I'm not a lawyer so I can't talk about whether it can be enforced in practice. But still, the fact that this open-source implementation doesn't reference the patent anywhere suggests that they didn't consider the legal implications properly.

[1] https://patents.google.com/patent/WO2014170646A1/fr?oq=WO%2f...

[2] https://whatfreewords.org/about.html

IANAL, and may be wrong about how US patent law works but AIUI:

- A patent isn't infringed unless an independent claim is infringed

- Claim #7 (the one with the equations) is a dependent claim (dependent on claim #1, which is independent)

The key questions seem to be:

- whether claim #1 is being infringed, and

- whether the patent is valid

This is exactly correct. The point about the formulas was meant as an illustrative example about the fact that the patent might well apply rather than an argumentative one.

It's a good choice of example... of why it might be invalid due to Alice.

It's open source. Good luck enforcing patent claims against something that can be copied endlessly for free. Ask how that turned out against ffmpeg.

Eh, if they pull all apps which use that library from apple's and google's app stores, that would be enough to make it useless.

Codec IP owners usually grant royalty-free rights to software video players.

They make money licensing IP to hardware manufacturers.

MPEG-LA does not. https://www.mpegla.com/programs/avc-h-264/ is their license scheme - while they allow a certain threshold below which cutoff isn't required, there's a lot of companies (such as Blizzard Entertainment!) whose primary distributed product is solely for video playback. H.264 requires a license for personal, at-home playback of encoded files. You have such a license if you have a Windows OS, an Apple OS, if you're using Chrome, or if you use a set-top box, or other "single-use" device which can play back H.264. You do not have such a license on linux with mplayer.

The FAQ strongly hints that they know there'll be some sort of legal blowback:

"Before WhatFreeWords existed, another group had released an open-source implementation of What3Words and published it on GitHub, and GitHub removed the repository after receiving a DMCA takedown notice from the What3Words company. It is likely that similar action would be taken against a WhatFreeWords repository on GitHub."

and:

"We prefer to remain anonymous. The What3Words company has shown that they will attempt to cause problems for people distributing open-source implementations of the What3Words geocoding system, so it is easier if we are anonymous."

Turns out that in order to evade DMCA action, they needed to do more than not being on github. As of yesterday I think..

"WhatFreeWords is currently unavailable due to DMCA action from the What3Words company"

What3Words is a horrible idea.

It's primary use is as a rent extraction device for its owners, if they can gain sufficient market share.

It is utterly useless without their app, uses a non-universal language (English vs math) and provides zero information about geography. The 3-word identification strings provide not the slightest clue whether two identified points are corners of the same building or literally on opposite sides of the planet (vs LatLong making it obvious that e.g., 42.36N 71.31W is relatively near 41.92N 71.48W).

It provides zero value to anyone trying to actually navigate, without using their app, and if you are going to navigate with an app, why not use any implementation of OpenStreetMap, or even the nearly ubiquitous Google or Apple maps? Hell, we're even better off with old paper postal maps than this scheme.

A slight about of temporary convenience for innumerate people is no good reason to implement a rent-extractive proprietary system.

They publish word lists for other languages.

https://support.what3words.com/en/articles/2212864-how-many-...

I don't think it is a particularly good idea, but insisting people use English isn't something they are doing.

Great -- so now we also have a translation issue! </sarc>

Nice that they're making a nod to avoiding language hegemony

However, the numerals used in LatLong navigation are about as universal as a language gets, requiring no translation in written communications, and are among the first words learned in second languages. In contrast, the three random words can fall subject to all kinds of translation issues whenever crossing a language barrier.

Makes me realize that we'd need not only their app, but also their official translation table; makes it seem even more like a fractally bad idea.

I commented last time on W3W coverage for French but it's beyond useless, they use very obscure words 90+% of native speakers have never heard of and you would need a dictionary even for well known addresses in Paris.

I wholeheartedly agree with your assessment.

In contrast, Google’s plus codes(https://plus.codes/) provide the same functionality but avoid the most of the pitfalls you mention.

If you wanted to encode the same data using words instead of alphanumeric characters (which is a debatable goal), then you could just combine it with WCodes (https://wcodes.org/)

Plus codes are also unencumbered. Their main disadvantage is being saddled, branding-wise, to the Google Plus fiasco.

Yes, just the general about page for the Google Open-Sourced Plus codes makes clear that they have thought of and solved at least a half dozen major issues ignored by the What3 group.

Just check the topics: Free, Accessible Offline, Easy to Use, Non-Exclusive, Independent of Borders, Identifiable, Works in Unmapped Places, Variable Precision, Useful of Navigation (e.g.,nearby places have similar codes). . .

I don't like the nature of it.

I went out walking and wild camping for a few days recently, in areas of not much signal. I usually check in with family as I move about when I do these things, with location and updated plans, and I was surprised that my brother recommended it to me. He said it's becoming quite popular in search and rescue applications.

As it goes I did what I normally do an just pin my location on WhatsApp when I get a bit of signal, but in the right (wrong) circumstances I could see it being a good way to quickly convey location.

I can also see numerous issues with it.

But it made me think maybe there are applications, beyond navigation where it might be useful.

I'll keep an eye on this project anyway, I hope it survives.

Here's a list of other alternatives to w3w

https://wiki.openstreetmap.org/wiki/What3words#Alternatives

The javascript implementation defines deg2rad as

  deg * Math.PI / 360

It's been a while since high school math, but shouldn't that be divided by 180? Looks like the numbers in the docs are based on this incorrect conversion.

You're right, thank you. This is a mistake. This will be corrected as quickly as possible.

ETA: This has been corrected. Thank you.

Sadly this is likely to get shut down by What3Words, who are very protective of their algorithm.

The authors know that: https://whatfreewords.org/faq.html

A subpoena will figure out who they are.

And then extradition orders in every country they're in?

You don't have to extradite the authors if they live in countries in which patent protection is strong and the legal systems are functional enough to impose due process upon them and enforce injunctions/damages. Extradition is typically a last resort, typically for criminal proceedings.

Moreover, an injunction is much more likely to be sought than damages, in which case extradition is unnecessary - all the patentees have to do is get an order from a judge forcing the hosts to take their code and site down.

Many countries don't recognize software patents, and the ones that do have some strict rules about what patents are valid. So, are you sure that this is a slam dunk? You don't seem to have any concept that it might be ruled invalid.

It’s a lot easier to accomplish this under the conditions I outlined above. That’s all I’m saying. People have gotten away with worse.

Many countries have strong patent protections and no software patents. If you wanted to say "have software patents" instead of "strong patent protections", it would make a bit more sense.

How's that? There are ways to post content anonymously: https://news.ycombinator.com/item?id=20220048

Why "sadly"? They know it is infringement. When you own a patent or trademark you HAVE to go after the infringers, otherwise you stop owning the patent.

This is not true in the US. There is a concept of abandonment for trademarks, but it's not that simple, and the concept does not apply to patents. There is a totally different thing called patent abandonment but it has nothing to do with going after infringers.

There is a statute of limitations on patent damages but you don't lose your patent for failing to enforce.

I'm not familiar with UK or European patent law but I expect it's similar in this regard.

Perhaps if they reside in, operate from or actively serve the US. In other parts of the world, software patents are unenforceable (as only implementations of ideas can be patented, not the ideas themselves).

Based on their wordlist, they do seem to use American English (so they are probably not located in the UK) but the authors can still operate this project from anywhere in the world.

Mapping some bits of data to a short set of words or pictures isn't even new, neither is the hashing step to ensure that similar coordinates don't use the same name. Validating (session) keys by hashing the key to generate a fingerprint and turning that into words (or these days, emoji) was invented before W3W was conceived. The patent itself is quite flimsy, considering it's just applying some formulas to do something that was already being done on basically any data.

Independently of the fact that I'm thoroughly against the use of proprietary algorithm for encoding, there's an interesting problem of the implementation of a range of similar problems to which somebody here maybe knows a nice effective solution:

(from their FAQ):

"The JavaScript library is about 2.4 megabytes. Almost all of this data is encoding data about how regions of the Earth are mapped to n values. There are over 300,000 regions. The raw data for the regions is 1.8 megabytes, and getting the data encoded in a JavaScript file raises this to 2.4 megabytes even after compression.

The Go library is about 2.6 megabytes of source code, for similar reasons, and produces executables of at least about 5.7 megabytes.

If you can shrink any of this data or make any other improvements, then please contact us and we will be pleased to use your improvements."

For the go executable to be so much larger than the source code, when he majority of that is data, there must be something going wrong...

Hello, I am Ujjwal - the inventor of WCode. Please checkout this page: https://wcodes.org/patents and https://wcodes.org/location And my webapp: https://location.wcodes.org

I have only ever heard of What3words in internet tech forums etc. Is it a well known company for the general public in some countries?

I wonder this too.

Honestly I have never seen it in use. In most countries, if something is on a street, it has a street address, which is more convenient than W3W.

I’m not sure how accurate “in most countries, if something is on a street, it has a street address” is. Even in USA, if my car breaks down while I’m going cross-country, I’d be nice to have a fast way to share location.

Would this not be possible though by sharing your lon/lat though? Plus easier to navigate to using gmaps or the like.

On the flip side, I have read stories of w3w success in giving an address to people where a normal street address is not possible, e.g. favelas https://www.reuters.com/article/us-innovation-what-3-words/w...

lat, long:

“I have fallen and broken my leg. I am under a rock at *thirty three point five nine six one three degrees north and one hundred and seventeen point seven nine six nine two degrees west. That’s thirty three point five nine six one three degrees north and one hundred and seventeen point seven nine six nine two degrees west.” Then wait for them to read it back to you. Then repeat that as many times as there are different people involved in the rescue.

Versus 3 word:

“I have fallen and broken my leg. I am under a rock at 3 word code: off-road — starfish — postmark.” Then wait for them to read it back to you. Then repeat that as many times as there are different people involved in the rescue.

Where I am rn, btw. (Though thankfully not my situation) And the number of decimal places more or less corresponds with the size of the 3 code and an appropriate level of precision, given the environment.

yes - very good point

https://www.bbc.com/news/uk-england-49319760

Apparently some emergency services recommend & use it.

No emergency service should use something patent-encumbered - but W3W markets to them all the time.

It seems that they have a pretty good PR department and had articles about them in a lot of major newspapers around the world. If had quite a few non-techy people ask me about W3W that I wouldn't have assumed would know about them.

I have only heard about it in articles complaining about it.

They pay automakers to build their engine into in-car nav systems. You can now enter w3w address into many cars directly, but not competing open systems like plus codes or xaddress or maidenhead or whatever.

I find the fact Steve Coast (the founder of OpenStreetMap) has worked for what3words pretty grotesque.

https://www.spatialsource.com.au/latest-news/founder-of-open...

It's a shame that he didn't have enough charisma and vision to become OSM's BDFL a la Python's Guido von Rossum. Having a single person could help transcend the current "continue status quo" and "lean OSM Foundation that does absolute minimum" paradigm which partially stem from from the fact that every change would meet with some opposition (which is hard to override if there's no clearly dominating power).

It's been shut down already.

Reply on news.ycombinator.com