5 ms·

I don't think I'm on board with this. Here are two names from the same street: NOZAXO-JEJEZU, NUZAXOJ-JAJEZUN and here's one from another location in the city

I don't think I'm on board with this. Here are two names from the same street: NOZAXO-JEJEZU, NUZAXOJ-JAJEZUN and here's one from another location in the city PABAXO-GAKEZU. First, I wouldn't like to prominently use these names on my business as I would be naming them nearly the same as any other one on my street. Second, it is not clear at a glance that the second address is in the same area. And without specifically saying "Atlanta" you lose all types of information about taxes, police, fire and other government type things which apply to areas bound by complex curves in lat-lon space. Here's three other addresses: BAJASU-BONEXO, DAFASU-DUTEXON, LIDEJA-KESERO Two of these are about 10 miles apart and the other is thousands of miles away. I can see some similarities to the ones close by but if just given a single UBI name, I doubt I'd have any clue where it is. It seems that actually seeing lat-lon numbers might make it easier to see that locations are close together. This is also ignoring streets, which carve complex paths through lat-lon space. I'm not sure if the street name is wrapped up in these UBI names but losing that information would be terrible for locals or honestly anyone else traveling. If you know something is on "broad st." and you are on broad st., then you can follow the road to get there. If you are looking for NUZAXOJ-JAJEZUN and you are at NOZAXO-JEJEZU, you may not realize that you are on the same street. I didn't go through the entire set of slides as it was pretty onerous, some having 7 clicks to reveal everything in them. I think I got the basic idea though and I don't really see a use case for this replacing any part of our current system. I would never name a business using these. They are unnatural and difficult to pronounce, so I'm not sure how the "Pine Hill" example is valid. Adding these to existing addresses would be useful, in the sense that adding lat-lon to an existing address would be useful but I don't find them more readable than lat-lon. And the UBIs have the negative effect of needing to be translated to lat-lon in order to find locations.

drostie 12y ago
That is the big take-home problem here. I looked up the address that they said was in Nigeria (it's actually in Mali), REJATE-POKAJA. I wanted to know, "well, what's the address of the building across the road?" and it turns out that that is: RIJATE-PIKAJAS. I think my first thought is, "Is that really what you want?" Those sound pretty darn similar, and it'd be easy to get confused. In reality you already have a lot of info about a place implicitly, so you want to focus on the stuff that you don't know, magnifying that to cover the whole number. It seems that they get something like 16-20 bits per word or so. I think my first approach in this case would be to feed lat/long through a cylindrical equal-area projection, since the math for those is pretty simple (so it can be implemented in diverse systems easily) and at least there is a semblance of fairness to the resulting numbers -- but really, I think what you want is to cover the Earth with a massive geodesic dome of points and just number the vertices with some algorithm, perhaps placing a great circle on the Greenwich line and then descending around in some sort of spiral. There's no fundamental reason I can see why we'd need to project at all, really. So, you form two numbers with the Gall-Peters projection: 2^9 * (most significant 9 bits of x) + most significant 9 bits of y 2^9 * (least significant 9 bits of x) + least significant 9 bits of y or else you label each of 2^18 geodesic faces in a spiral and do coordinates on the faces; either way. You take these numbers, and multiply them by some large prime, modulo the number that you've got. Prime multiplication is a simple way to permute the numbers so that adjacent vertices are unlikely to be numbered similarly. If you really wanted, you could then XOR it with some fixed pattern and multiply by another large prime, which may help with names being similar at the poles themselves or something (depending on the size of the prime). Any quasi-random permutation will work, I just figure "let's take the ones that are only a couple fixed lines of code." The nice thing about these permutations, too, is that they're invertible: someone gives you a name, you convert it to a number, you pass it back through this, and you get map/geodesic coordinates. Then you convert those numbers into syllables. This way you can know that your city is mostly DAFAXO except the northeastern part is part of LINUZU, completely different names. The most significant bits are thus scrambled but isolated. The least significant bits are scrambled separately and give every point on the ground a wildly different name, so that nobody confuses your place with one down the street.
- roberdam 12y ago
  Hi Drostie, It is an excellent analysis, when designing the algorithm there are 2 choices, to make the Global Addresses somehow related, so you can know that REJATE-POKAJA is near RIJATE-PIKAJAS, or to make them totally different so you wont get confused by the similarity, both approaches have their advantages and disadvantages , and as you say, is not something difficult to implement
roberdam 12y ago
Hi Wuliwong and thanks for your comments is a great feedback, Im Robert the developer behind the idea. Location Aware Names used as Global Addresses is one of the choices you can use. If you want to keep the current Street & Number format you can use Location Aware Street Names, so your business address can be at street "Clerigo Tigorros 155" and the building in front of you will be at "Clerigo Tigorros 158", you can use it just like any other street with the Street name and number. You can get there using your offline address like "Clerigo Tigorros 155@gazado-fifene" or even if the town is already on a map using "Clerigo Tigorros 155@(Naranjito, itapua, Paraguay)" so every street in town will share the same Zone code @gazado-fifene if you want to know if they are in the same area. http://ubicate.me/street/index?utf8=%E2%9C%93&street=Clerigo+Tigorros+155%40gazado-fifene&commit=Search&streetlist= http://ubicate.me/street/index?utf8=%E2%9C%93&street=Clerigo... In case you wonder this name was generated for a spanish speaking town, clerigo means "cleric", so you can use any language you want to generate the street names. I made a non animated and "cut to the chase" slideshow you can acces on a "Slides are too slow?" link now on the website, hope that improves the experience!. If you have any other question or suggestion let me know.

drostie 12y ago

That is the big take-home problem here. I looked up the address that they said was in Nigeria (it's actually in Mali), REJATE-POKAJA. I wanted to know, "well, what's the address of the building across the road?" and it turns out that that is: RIJATE-PIKAJAS. I think my first thought is, "Is that really what you want?" Those sound pretty darn similar, and it'd be easy to get confused. In reality you already have a lot of info about a place implicitly, so you want to focus on the stuff that you don't know, magnifying that to cover the whole number. It seems that they get something like 16-20 bits per word or so. I think my first approach in this case would be to feed lat/long through a cylindrical equal-area projection, since the math for those is pretty simple (so it can be implemented in diverse systems easily) and at least there is a semblance of fairness to the resulting numbers -- but really, I think what you want is to cover the Earth with a massive geodesic dome of points and just number the vertices with some algorithm, perhaps placing a great circle on the Greenwich line and then descending around in some sort of spiral. There's no fundamental reason I can see why we'd need to project at all, really. So, you form two numbers with the Gall-Peters projection: 2^9 * (most significant 9 bits of x) + most significant 9 bits of y 2^9 * (least significant 9 bits of x) + least significant 9 bits of y or else you label each of 2^18 geodesic faces in a spiral and do coordinates on the faces; either way. You take these numbers, and multiply them by some large prime, modulo the number that you've got. Prime multiplication is a simple way to permute the numbers so that adjacent vertices are unlikely to be numbered similarly. If you really wanted, you could then XOR it with some fixed pattern and multiply by another large prime, which may help with names being similar at the poles themselves or something (depending on the size of the prime). Any quasi-random permutation will work, I just figure "let's take the ones that are only a couple fixed lines of code." The nice thing about these permutations, too, is that they're invertible: someone gives you a name, you convert it to a number, you pass it back through this, and you get map/geodesic coordinates. Then you convert those numbers into syllables. This way you can know that your city is mostly DAFAXO except the northeastern part is part of LINUZU, completely different names. The most significant bits are thus scrambled but isolated. The least significant bits are scrambled separately and give every point on the ground a wildly different name, so that nobody confuses your place with one down the street.

roberdam 12y ago
Hi Drostie, It is an excellent analysis, when designing the algorithm there are 2 choices, to make the Global Addresses somehow related, so you can know that REJATE-POKAJA is near RIJATE-PIKAJAS, or to make them totally different so you wont get confused by the similarity, both approaches have their advantages and disadvantages , and as you say, is not something difficult to implement

roberdam 12y ago

Hi Wuliwong and thanks for your comments is a great feedback, Im Robert the developer behind the idea. Location Aware Names used as Global Addresses is one of the choices you can use. If you want to keep the current Street & Number format you can use Location Aware Street Names, so your business address can be at street "Clerigo Tigorros 155" and the building in front of you will be at "Clerigo Tigorros 158", you can use it just like any other street with the Street name and number. You can get there using your offline address like "Clerigo Tigorros 155@gazado-fifene" or even if the town is already on a map using "Clerigo Tigorros 155@(Naranjito, itapua, Paraguay)" so every street in town will share the same Zone code @gazado-fifene if you want to know if they are in the same area. http://ubicate.me/street/index?utf8=%E2%9C%93&street=Clerigo+Tigorros+155%40gazado-fifene&commit=Search&streetlist= http://ubicate.me/street/index?utf8=%E2%9C%93&street=Clerigo... In case you wonder this name was generated for a spanish speaking town, clerigo means "cleric", so you can use any language you want to generate the street names. I made a non animated and "cut to the chase" slideshow you can acces on a "Slides are too slow?" link now on the website, hope that improves the experience!. If you have any other question or suggestion let me know.