Which is a big problem. They want to own the address of most of the people in the world. They might get away with it, too; their word list can be copyrighted.
"Where the streets have no name," sang rock frontman Bono on one of U2's biggest hits.
"It must be hell being a postman, then," came the sarcastic rejoinder from the music press.
But the system they come up with is still hell for a postman. He either remembers the 111111 three-word combinations for his square kilometer of delivery, or he can't work without an electronic device.
I see the problem, but hope that there are better solutions. Especially ones that are free: A non-free address system is absurd in my opinion, to put it mildly.
Well, if you ever try providing clean water and medical services to one of these places, you will quickly find that your first challenge is distributing resources in an efficient, non-wasteful way.
And to do that, you need to take a census and find out how many people live in each area, how many are lacking service, and to what extent.
And to do that, you need to have some way of counting people that both humans and computers can understand. Which brings us back to this very elegant address system.
The latitude and longitude the words map to are better solutions since at least they correspond to the direction the postman is moving in a logical way.
They might be long strings of digits, but so are telephone numbers
Nice idea but there is also sacrifice of the relativity of the places. Based on the calculations, looks like the order of the words are important too.
"With 40,000 dictionary words, you have 64 trillion combinations, and there are 57 trillion squares."
If the dictionary words can be doubled, the order of the words can be made unimportant too. That way it would have ~80 Trillion combinations (64 Trillion/3!) * 2^3.
I may be missing something here, but it seems like this solves only half of the problem. Sure it's helpful to cover the earth in 3m squares with unique names, but in order to actually deliver something you need to be able to navigate to it.
The advantage of a (building number, street name, locality) address is that it identifies some point on the edge of a graph, and that the edges and vertices are well-known.
I'll grant that within any locality, people generally know how to get from point A to point B (and frequently don't know the names of the streets), but it seems like what you really need is a way to map the unique 3m squares onto that knowledge.
Even with that problem solved, another problem comes up with planning deliveries efficiently. There's no information embedded in the what3words address that encodes any notion of adjacency or relative distance. I realize that this is probably intentional to reduce the likelihood of errors in addresses going unnoticed, but it means determining nearness requires access to the whole mapping data set (or at least enough to determine that a what3words address isn't in the set of ones you care about).
That is partially done with Geohash (http://geohash.org/) and other variants. The actual step here is usage of dictionary words as digits for encoding instead of a bunch of random characters.
Alternatively, such a system can be done by interleaving bitstrings for coordinates then converting the result to base 10000 or something and encoding each digit with a word. By using that you will get a little bit of hierarchy (the coordinates starting with same digits will correspond to sequences starting with same word). In this way we will know which zone starts with e.g. chocolate.kit.photo then it will be easier to navigate. Of course it won't beat a graph-based solution like street and building numbers.
I like how my home/property has ~70 addresses and none are similar. To use this I'll need to pick one and make sure its used exclusively, otherwise I'd not only confuse the postman, I'd confuse anyone I need to give location to.
Without organization or structure of the words I don't see how it's helpful. Knowing the destination and my latitude and longitude, at least I'd know I was close, without having to reference this service.
An new address solution should solve multiple issues, utilizing existing knowlege, without making it more difficult.
Logical organization should be included. For example, we typically know where England or Canada is, or know the difference between SanDiego.California or adrar.algeria and subdivide from there. At least at this point, most people would know a general area without having to reference forgot.paved.ruler from an external source.
"Hey meet me at forgot.paved.ruler, it's just down the way from mixers.gulls.sing, or was that sings?.." Forgetting that 's' puts you on another continent, but it's close enough?
Geohash is great, but doesn't include any error detection or correction. This means that one (undetectable) symbol changes, and you could be half-way around the world from the original Geohash location.
The EZCOD encoding provides accuracy within 3 meters AND error detection/correction with 10 symbols. For example, where my daughter was born: R3U 1JU QUY.0 (try it at http://ezcod.com)
If you want 20mm accuracy AND 5-nines certainty of correctness, you can have that in 15 symbols: R3U 1JU QUY L02.XJ8
Available in C++, Javascript (via Emscripten, production-ready and CDN hosted), Python (via Swig) and REST APIs:
Use EZCOD for free. Forever. For whatever application you want. Available under either GPLv3 or completely free Commercial licensing, at your preference.
Coming from a country with a working, comprehensive, addressing system, I never understood how special this was until I started extensively traveling. It's kind of mind-boggling how unique of an invention addresses are and yet how simple they are.
It requires only two things
- give an identifier to each road, numbers are perfectly fine (the scheme isn't even all that important)
- each property plot along a given road gets an incrementing number, plots with multiple housing units are given some kind of subdivider (Apartment, Unit, etc.). This works best with oddly numbered properties on one side and evenly numbered properties on the other.
So long as you use unique identifiers, you don't even need states, counties, zip codes or other administrative districts -- though this makes things easier -- all you need to know if the block of property numbers the address lives in.
An alternate approach is to simply subdivide a territory into smaller and smaller zones until you get to a small enough spot on the planet that you've uniquely identified a plot of land or a building or whatever. I believe Korean addresses used to use something like this but are slowly moving on to a mixed mode system like the U.S. uses (subdivide areas, then use street names and numbers).
It's also seems absolutely insane that the concept of an address, or parts of an address can be some kind of Intellectual Property that a company holds some kind of right to. My understanding is that this used to be a significant problem with addressing in the U.K. and remains one in Ireland. The last time I visited Ireland, outside of the cities, addresses were give to us as Lat,Long coordinates! Absolute insanity.
I remember trying to send a package to a friend in rural England once, the address was something like "The Yellow Home behind the Red Barn, Essex, U.K.". It took 8 weeks to arrive, I have no idea how much of that was spent by the Royal Mail matching color swatches and building combinations.
It's an interesting notion that we couple postal service to addresses. In some places, like Dubai, they are decoupled and the post only makes its way to a P.O. Box, not to a residence -- which of course have no address, and end up with descriptions like "Building ABC Behind the Market, off of the Second Exit, 2 Lefts and a Right after entry" which is necessary for non-government postal services like Fedex. Oddly, this system makes a kind of sense in the Internet age, unless you actually have to get somewhere.
In some countries, local landmarks and these sorts of quasi-turn-by-turn directions-as-addresses are the norm. Oddly, they often persist long after what they describe is gone. "The Red House behind the Temple" might be somebody's address, but the Temple fell during an Earthquake 20 years before and the house is no longer Red. The claim is that the local postman has enough local knowledge to make this work, but there's so many obvious failure modes to make this sort of "system" seem absurd.
If you look back long enough, the U.S. didn't always have a useful system. Old Revolutionary War era advertisements look like the kinds of semi-useless addresses you see elsewhere. Addresses definitely appears to be product of cities, but it's useful when it's used on a national scheme. It's so simple and so absurdly effective, it's always seemed bizarre to me that it's just not universal.
This proposal is interesting, but ultimately a bad idea. Correct addressing encodes both a unique location and navigation way points.
Take a U.S. address:
1234 Main St.
Beverly Hills, CA 90210
Beverly Hills, CA is honestly not necessary, but does make it human readable, 90210 tells you this location is out west and the post office has a map that subdivides it even more to a specific delivery zone.
There's hopefully only one Main St. in that zip code, and the building is on the 1200 block of that street on the even side. When you arrive th...
I'd agree with you to some extent but postcodes (the UK equivalent of ZIP codes) were actually introduced before ZIP was in the US. If you were given an address in 'rural England' like that at any point in the past >40 years then it was most likely the addressees choice rather than the actual postal system. The postcode also makes everything apart from the number irrelevant in the eyes of the post office - you can address items to 10, GL10 2AB and it will end up where it needs to go.
That's good to know. I believe there was a post code (this exchange happened about 10 years ago). I recall there's been several interesting exchanges over rights access to the postal code locations in the U.K. as internet mapping started to become more prevalent. I think this finally changes ~2009/2010? At least it finally happened, but I cannot fathom the thought process that would have made a campaign necessary to force the release of basic location data.
> It took 8 weeks to arrive, I have no idea how much of that was spent by the Royal Mail matching color swatches and building combinations.
Most probably none. The Royal Mail has solved addressing but the Border Agency has not solved customs. In-country mail can be guaranteed for same-day delivery; imports - especially from the US - take weeks to process.
How about retrofitting DNS to do this? Encode a persons GPS coordinates in a TXT record. Now instead of writing my address on an envelope, I write me.example.com. The mailman/post office can then look up someone's coordinates via a DNS query. The benefit of this is that when I move, I can simply update my DNS entry. I could either host my own address or 3rd party could. Local government could grid up their own regions and assign address to fixed locations. You could choose to CNAME your address to the government allocated address or use the government issued one for the plot.
One critical issue in this (apart from the proprietary approach) might be the mapping of adresses between languages, since a random Indian or Chinese person will probably not benefit from English words.
Can you identify subsets consisting of 10,000s of words in different languages that have one-to-one maps to any other global language subset for w3w? (i.e. without multiple words being suitable for a translation of the word)
I lived in Saipan for a few months, part of the CNMI. A couple decades ago they attempted assigning road names and addresses, but what you find is that almost no one uses them. They don't have a postal delivery infrastructure, so those who can afford it get private company PO Boxes (the official PO boxes are extremely limited).
To me it seems like the real challenge is funding and setting up a postal infrastructure. Creating 'addresses' of any scheme would emerge organically out of necessity from that.
> Sure it's helpful to cover the earth in 3m squares with unique names ...
I'm actually not convinced. Squares don't tessellate a sphere; you have singularities at the poles. Sure, there aren't many people living there, but it's important to think about the limitations of this approach.
If you think this is really just an edge case, remember that projecting a square onto a sphere necessitates some distortion. You can choose to distort angles or distances, but you have to choose one. For distances, well, a "3m x 3m" square becomes essentially meaningless. The problem gets worse when you increase latitude. Canada, northern Europe and Russia are especially prone to this problem. The same argument applies to locations in the Southern hemisphere as well.
This is how I would do it. Your base unit is a circle with radius 180cm, which is approximately 10m^2.
Place one circle at the north pole. Your starting angle is 0 degrees longitude. Place a circle at the south pole, and slide it along that line of longitude until your sliding circle touches any other already-placed circle. Then "roll" the circle to the local minimum distance between its center and the north pole; whenever moving it an epsilon east or west would not allow the center to get closer, stop. Add 222.5 degrees (actually (sqrt(5)-1)*180), modulo 360. Slide in another circle. Add 222.5 degrees, modulo 360. Slide in another circle. Repeat until the circle placed at the south pole intersects an already-placed circle.
Move each center point to the nearest grid-based point at a given precision, then name them. If you stand on the named point, you should be able to touch anything using that point as its address, without moving your feet.
A 10m^2 unit cell is a very human-sized measurement, but you are correct that forcing them into an orthographic grid is pointless when the addresses are names and not numbers.
There's no way any living human could memorize the name-location mapping for all addresses, or even just all the land-based addresses. People would need to look them up in an index from a numeric coordinate, and vice versa. As such, there is no need to orient the address points to any existing coordinate system.
That's true for large chunks, but I'd imagine 3 meters is small enough to remain useful well past the 85 degrees latitude mark. Straight Dope informs me that the circumference up there is approximately 24k miles, which becomes 38,624,256 meters which means approximately 1 million divisions.
Dammit. I thought I had been careful enough to get the correct number.
Alright. Picking the simplest formula off Wikipedia, cosine of 85 degrees is ~0.0871557 times a (6378137m) times pi is ~1746383.9m for the circumference. That's still approximately 500k squares at the exact latitude. My point is one of the sheer magnitude still available.
What I should do is calculate how many decimal points down from 90 degrees 3 meters is and get that circumference, since that's the low bound, but I'm not really feeling up to the number-plugging.
I could be incorrect about what a is; I presumed it is the semi-major axis on WGS84, but I might have picked the wrong thing.
...
Ooh. I forgot there's a Wolfram|Alpha. ... And 10 minutes later I haven't convinced the input parsing to give me what I want. Oh well.
This seems as behind-the-curve as printing out email and delivering it to physical mailboxes. As long as you can accept delivery of electronic documents, nobody should care where you "reside." If you need to take delivery of goods, you can specify a designee for that purpose, as and where needed.
I have to think "ugh" when I read about this idea. Random three words to every square seems like chaos to me.
These also take a lot of bytes to store. In the US, every address or "delivery point" can be represented by 11 numerals: zip+4+2. If you read the barcodes on your received US mail, you will see the normally invisible +2. (zip+4 is usually granular to one side of a block or MDU mailbox cluster, and the extra two numerals represent a unique mailbox within the zip+4.)
Amateur radio people have organized the world into grid squares, which are a bit larger as they are minutes of latitude/longitude in either dimension. However they are like postcodes, in that once you are familiar with the system you can glance at one and have a rough idea of whether it is near you or not, and where on the map it is. For instance, the White House in Washington is within grid square FM18lv.
Introduction to the grid square system at [1] and a rudimentary mapping tool at [2].
I'm not sure if there is any reason why the grid squares couldn't be further subdivided. Even then, this wouldn't solve the problem in some places I've been: The barrios around Caracas are not only tightly packed in two dimensions, but they have been building their homes on top of each other, so a 3m x 3m square could very well contain five homes or more.
I recently learned that Northern Canada has this problem.
Most of the communities in the Arctic do not have street names. Canada Post has been forcing them in the last few years to start, so, for example, a community called Aklavik officially has one street, Aklv St. Even though there are physically more streets than that, and streets at right angles, they're all called Aklv St.
I work for the telco that services these communities, and it's hilarious to look at the street addresses in our plant and provisioning systems - addresses are literally "brown bld across from red bld", "House near lake" etc.
Not surprisingly, I'm having trouble writing code to parse street addresses :)
Oh, classic address trap. I feel for you, my friend. It makes you wonder about introducing a new address system that takes into account your actual location in the globe (with GPS). Not a trivial undertaking itself. I assume it has been thought off before. Anyone have more info on this type of setup?
After visiting a friend in Costa Rica a few years ago, I asked his address as we were leaving so I could send him stuff from the states. It was "800m up from the cemetery in [the town below his house on the mountain]".
I've lived in the Dominican Republic for several years. Addresses in some communities here tend to be like, "Over by Carlos' house where Maria's grandma used to live."
An imprecise, centralized, English-centric system for labeling the planet. Why not just use GPS coordinates? It's much more universal, distributed and precise.
This article is clearly placed by the guy featured in it whose company is promoting an alternative to street addresses.
The article incorrectly states that people without street addresses have no postman and can not get mail.
This is not the case at all. I've lived in many places throughout the world without street addresses. Mail is delivered to addresses consisting of the person's name and things such as their village, precinct or neighborhood.
The articles assumption and promotion of non-factual claims as true suggests the author and the company simply do not know what they are talking about.
40 comments
[ 3.3 ms ] story [ 110 ms ] threadIn the Rio example, a postal service or courier would have to pay w3w to license the words->lat/lon lookup before they could deliver anything.
Really? Don't they have connections with Ogilvy and Mather?
"Where the streets have no name," sang rock frontman Bono on one of U2's biggest hits. "It must be hell being a postman, then," came the sarcastic rejoinder from the music press.
But the system they come up with is still hell for a postman. He either remembers the 111111 three-word combinations for his square kilometer of delivery, or he can't work without an electronic device.
I see the problem, but hope that there are better solutions. Especially ones that are free: A non-free address system is absurd in my opinion, to put it mildly.
Access to clean water and basic medical services will do far more.
And to do that, you need to take a census and find out how many people live in each area, how many are lacking service, and to what extent.
And to do that, you need to have some way of counting people that both humans and computers can understand. Which brings us back to this very elegant address system.
They might be long strings of digits, but so are telephone numbers
"With 40,000 dictionary words, you have 64 trillion combinations, and there are 57 trillion squares."
If the dictionary words can be doubled, the order of the words can be made unimportant too. That way it would have ~80 Trillion combinations (64 Trillion/3!) * 2^3.
The advantage of a (building number, street name, locality) address is that it identifies some point on the edge of a graph, and that the edges and vertices are well-known.
I'll grant that within any locality, people generally know how to get from point A to point B (and frequently don't know the names of the streets), but it seems like what you really need is a way to map the unique 3m squares onto that knowledge.
Even with that problem solved, another problem comes up with planning deliveries efficiently. There's no information embedded in the what3words address that encodes any notion of adjacency or relative distance. I realize that this is probably intentional to reduce the likelihood of errors in addresses going unnoticed, but it means determining nearness requires access to the whole mapping data set (or at least enough to determine that a what3words address isn't in the set of ones you care about).
I guess the main thing here is that we can now express "GPS coordinates" in easy to remember and human-friendly phrases.
Alternatively, such a system can be done by interleaving bitstrings for coordinates then converting the result to base 10000 or something and encoding each digit with a word. By using that you will get a little bit of hierarchy (the coordinates starting with same digits will correspond to sequences starting with same word). In this way we will know which zone starts with e.g. chocolate.kit.photo then it will be easier to navigate. Of course it won't beat a graph-based solution like street and building numbers.
An new address solution should solve multiple issues, utilizing existing knowlege, without making it more difficult.
Logical organization should be included. For example, we typically know where England or Canada is, or know the difference between SanDiego.California or adrar.algeria and subdivide from there. At least at this point, most people would know a general area without having to reference forgot.paved.ruler from an external source. "Hey meet me at forgot.paved.ruler, it's just down the way from mixers.gulls.sing, or was that sings?.." Forgetting that 's' puts you on another continent, but it's close enough?
https://news.ycombinator.com/item?id=8614198
It's already implemented in mongo, elastic, solr, etc, supports arbitrary length/precision codes, adjacent cells have similar hashes. http://www.movable-type.co.uk/scripts/geohash.html http://mapzen.github.io/leaflet-spatial-prefix-tree/
The EZCOD encoding provides accuracy within 3 meters AND error detection/correction with 10 symbols. For example, where my daughter was born: R3U 1JU QUY.0 (try it at http://ezcod.com)
If you want 20mm accuracy AND 5-nines certainty of correctness, you can have that in 15 symbols: R3U 1JU QUY L02.XJ8
Available in C++, Javascript (via Emscripten, production-ready and CDN hosted), Python (via Swig) and REST APIs:
http://hardconsulting.com/products/13-reed-solomon#EZCOD
https://github.com/pjkundert/ezpwd-reed-solomon
http://ezcod.com
Use EZCOD for free. Forever. For whatever application you want. Available under either GPLv3 or completely free Commercial licensing, at your preference.
It requires only two things
- give an identifier to each road, numbers are perfectly fine (the scheme isn't even all that important)
- each property plot along a given road gets an incrementing number, plots with multiple housing units are given some kind of subdivider (Apartment, Unit, etc.). This works best with oddly numbered properties on one side and evenly numbered properties on the other.
So long as you use unique identifiers, you don't even need states, counties, zip codes or other administrative districts -- though this makes things easier -- all you need to know if the block of property numbers the address lives in.
An alternate approach is to simply subdivide a territory into smaller and smaller zones until you get to a small enough spot on the planet that you've uniquely identified a plot of land or a building or whatever. I believe Korean addresses used to use something like this but are slowly moving on to a mixed mode system like the U.S. uses (subdivide areas, then use street names and numbers).
It's also seems absolutely insane that the concept of an address, or parts of an address can be some kind of Intellectual Property that a company holds some kind of right to. My understanding is that this used to be a significant problem with addressing in the U.K. and remains one in Ireland. The last time I visited Ireland, outside of the cities, addresses were give to us as Lat,Long coordinates! Absolute insanity.
I remember trying to send a package to a friend in rural England once, the address was something like "The Yellow Home behind the Red Barn, Essex, U.K.". It took 8 weeks to arrive, I have no idea how much of that was spent by the Royal Mail matching color swatches and building combinations.
It's an interesting notion that we couple postal service to addresses. In some places, like Dubai, they are decoupled and the post only makes its way to a P.O. Box, not to a residence -- which of course have no address, and end up with descriptions like "Building ABC Behind the Market, off of the Second Exit, 2 Lefts and a Right after entry" which is necessary for non-government postal services like Fedex. Oddly, this system makes a kind of sense in the Internet age, unless you actually have to get somewhere.
In some countries, local landmarks and these sorts of quasi-turn-by-turn directions-as-addresses are the norm. Oddly, they often persist long after what they describe is gone. "The Red House behind the Temple" might be somebody's address, but the Temple fell during an Earthquake 20 years before and the house is no longer Red. The claim is that the local postman has enough local knowledge to make this work, but there's so many obvious failure modes to make this sort of "system" seem absurd.
If you look back long enough, the U.S. didn't always have a useful system. Old Revolutionary War era advertisements look like the kinds of semi-useless addresses you see elsewhere. Addresses definitely appears to be product of cities, but it's useful when it's used on a national scheme. It's so simple and so absurdly effective, it's always seemed bizarre to me that it's just not universal.
This proposal is interesting, but ultimately a bad idea. Correct addressing encodes both a unique location and navigation way points.
Take a U.S. address:
1234 Main St. Beverly Hills, CA 90210
Beverly Hills, CA is honestly not necessary, but does make it human readable, 90210 tells you this location is out west and the post office has a map that subdivides it even more to a specific delivery zone.
There's hopefully only one Main St. in that zip code, and the building is on the 1200 block of that street on the even side. When you arrive th...
That's good to know. I believe there was a post code (this exchange happened about 10 years ago). I recall there's been several interesting exchanges over rights access to the postal code locations in the U.K. as internet mapping started to become more prevalent. I think this finally changes ~2009/2010? At least it finally happened, but I cannot fathom the thought process that would have made a campaign necessary to force the release of basic location data.
Most probably none. The Royal Mail has solved addressing but the Border Agency has not solved customs. In-country mail can be guaranteed for same-day delivery; imports - especially from the US - take weeks to process.
[1] https://tools.ietf.org/html/rfc1876
Can you identify subsets consisting of 10,000s of words in different languages that have one-to-one maps to any other global language subset for w3w? (i.e. without multiple words being suitable for a translation of the word)
To me it seems like the real challenge is funding and setting up a postal infrastructure. Creating 'addresses' of any scheme would emerge organically out of necessity from that.
I'm actually not convinced. Squares don't tessellate a sphere; you have singularities at the poles. Sure, there aren't many people living there, but it's important to think about the limitations of this approach.
If you think this is really just an edge case, remember that projecting a square onto a sphere necessitates some distortion. You can choose to distort angles or distances, but you have to choose one. For distances, well, a "3m x 3m" square becomes essentially meaningless. The problem gets worse when you increase latitude. Canada, northern Europe and Russia are especially prone to this problem. The same argument applies to locations in the Southern hemisphere as well.
Place one circle at the north pole. Your starting angle is 0 degrees longitude. Place a circle at the south pole, and slide it along that line of longitude until your sliding circle touches any other already-placed circle. Then "roll" the circle to the local minimum distance between its center and the north pole; whenever moving it an epsilon east or west would not allow the center to get closer, stop. Add 222.5 degrees (actually (sqrt(5)-1)*180), modulo 360. Slide in another circle. Add 222.5 degrees, modulo 360. Slide in another circle. Repeat until the circle placed at the south pole intersects an already-placed circle.
Move each center point to the nearest grid-based point at a given precision, then name them. If you stand on the named point, you should be able to touch anything using that point as its address, without moving your feet.
A 10m^2 unit cell is a very human-sized measurement, but you are correct that forcing them into an orthographic grid is pointless when the addresses are names and not numbers.
There's no way any living human could memorize the name-location mapping for all addresses, or even just all the land-based addresses. People would need to look them up in an index from a numeric coordinate, and vice versa. As such, there is no need to orient the address points to any existing coordinate system.
As long as the system works, what's the big deal?
Alright. Picking the simplest formula off Wikipedia, cosine of 85 degrees is ~0.0871557 times a (6378137m) times pi is ~1746383.9m for the circumference. That's still approximately 500k squares at the exact latitude. My point is one of the sheer magnitude still available.
What I should do is calculate how many decimal points down from 90 degrees 3 meters is and get that circumference, since that's the low bound, but I'm not really feeling up to the number-plugging.
I could be incorrect about what a is; I presumed it is the semi-major axis on WGS84, but I might have picked the wrong thing.
...
Ooh. I forgot there's a Wolfram|Alpha. ... And 10 minutes later I haven't convinced the input parsing to give me what I want. Oh well.
These also take a lot of bytes to store. In the US, every address or "delivery point" can be represented by 11 numerals: zip+4+2. If you read the barcodes on your received US mail, you will see the normally invisible +2. (zip+4 is usually granular to one side of a block or MDU mailbox cluster, and the extra two numerals represent a unique mailbox within the zip+4.)
Amateur radio people have organized the world into grid squares, which are a bit larger as they are minutes of latitude/longitude in either dimension. However they are like postcodes, in that once you are familiar with the system you can glance at one and have a rough idea of whether it is near you or not, and where on the map it is. For instance, the White House in Washington is within grid square FM18lv.
Introduction to the grid square system at [1] and a rudimentary mapping tool at [2].
I'm not sure if there is any reason why the grid squares couldn't be further subdivided. Even then, this wouldn't solve the problem in some places I've been: The barrios around Caracas are not only tightly packed in two dimensions, but they have been building their homes on top of each other, so a 3m x 3m square could very well contain five homes or more.
[1]: http://www.arrl.org/grid-squares [2]: http://www.qrz.com/gridmapper
Most of the communities in the Arctic do not have street names. Canada Post has been forcing them in the last few years to start, so, for example, a community called Aklavik officially has one street, Aklv St. Even though there are physically more streets than that, and streets at right angles, they're all called Aklv St.
I work for the telco that services these communities, and it's hilarious to look at the street addresses in our plant and provisioning systems - addresses are literally "brown bld across from red bld", "House near lake" etc.
Not surprisingly, I'm having trouble writing code to parse street addresses :)
The article incorrectly states that people without street addresses have no postman and can not get mail.
This is not the case at all. I've lived in many places throughout the world without street addresses. Mail is delivered to addresses consisting of the person's name and things such as their village, precinct or neighborhood.
The articles assumption and promotion of non-factual claims as true suggests the author and the company simply do not know what they are talking about.