ronny, First, I have to say you know what you are doing (so apologies if my message below doesn't make sense, it's friday, I'm tired, and just having beer right now ;))
1) What I have in mind right now is how do we make sense (and map for OSM) existing addresses. I drafted the proposed schema (based on Karsluche) to accommodate existing address variations. In addition, I would like for us to create a web interface for others to easily add addreses to OSM. I see something like openstreetbugs' interface. We can add several textboxes where anybody can add an address. This can then be exported as a gpx or osm file. An OSM mapper can then add to the main dbase or verify the address on the ground. This way, non-OSMer can easily add addresses. Anybody willing to create a similar web-based app? We can glue in several codes from osm, openlayers and some back-end dbase. 2) On your proposal to create a community-driven address/geocoding/postcode schema, I raise my hand to contribute in developing this (although I am more of mapper than a coder). I see all the benefits of doing this. Several insights on the proposal: - We have to ensure "cross-compatibility" of the new schema to existing address variations in the Philippines and to OSM - It has to be applicable in both rural and urban areas - Open-license both the data and the generated algorithm for searching and geocoding Finally, I suggest we start rolling this out ASAP. We can get others to support the proposed design if we have a working example in place for testing and evaluation. On Thu, Mar 19, 2009 at 7:42 PM, Ronny Ager-Wick - Develo Ltd. <r...@develo.ltd.uk> wrote: > Maning and Rally, I am very happy that you mentioned this problem! > > I have worked with addressing systems for years. Having made database > schemas and software for direct marketing companies, mail order businesses > and a world wide location search system, I have considerable expertise in > this area. I have also been annoyed with the complete mess the Philippine > addressing "system" is in, and have been wanting to do something with it. > This may be the kick I need to get going. > Addressing systems look so easy when you just glance over them, but when you > dig deeper into the matter, like Rally has done here, you discover that it's > immensely complex. If the database schema holding this information is not > designed well, it's going to be a complete mess forever after. > I have analysed the situation in some other countries and found their > weaknesses: > > Norway (where I'm from): Postal codes are under the control of the postal > and telecommunications authority (which means government). This is the first > mistake. Anyone who has dealt with the government knows that thay are > utterly incapable of doing anything right. Result: small towns have one post > code, similar to the phils. Bigger (or randomly selected small towns) have > numerous postal codes which are assigned so they fit the routes of the local > post office! Completely useless to anyone but themselves... > Norway is also complicated, but not as messy as the Philippines. The most > complex addresses can be like this one: > Leil 42, 2. etg, oppg. C > Storebakken 12 > 0123 OSLO > The least complex one can be like the one of my Grand parents before: > 4760 BIRKELAND > (that's it) > The reason for this is that it's a small village, they live by the main > road, which has no name (only a road number which never is used in an > address) and the one delivering mail there knows everybody anyway. This is > still a valid address. > The good thing about Norway is that the street name is unique within the > postal code - in fact within the municipality (which can have 1 or more > postal codes as explained above). This luxiru we don't have in the Phils. > > Moving on the the UK, which has an excellent postal code system - in my > opinion even better than the US. Not perfect, but a good model. > Their "post codes" consists of four sections: > Postcode area > Postcode district > Sector > Unit > Examples: > GU35 9NQ - GU = Guildford (the nearest big town), 35=district within the > Guildford area, 9 = Sector, NQ = Unit > W2 6DB - W = West London, 2 = district, 6 = sector, DB = unit > EC2R 8AY - EC = East Central London, 2R = district, etc. > See http://en.wikipedia.org/wiki/UK_postcodes for full details. > The addresses in the UK are also structured in a way: > For example, my address is: > Flat 12 > Cleveland House > 55 Cleveland Square > London > W2 6DB > Address explained: > Flat 12 = flat number, apartment number, room number, etc (section of > building). Normally not included in address finder software. > Cleveland House - House name. in this case redundant information, as house > number is specified below. A UK address can have a house name OR house > number, or both. > 55 Cleveland Square - House number + street name > London - town/city (London is special as there is no need for county) > W2 6DB - postcode > We used to rent a farm house earlier, the address was: > Sandybridge Farm, Main Road > Kinglsey, Bordon, Hampshire > GU35 9NQ > Address explained: > Sandybridge Farm, Main Road - house name + street name (basically the same > as house number + street name) > Kinglsey, Bordon, Hampshire - Area, Town, County > GU35 9NQ - postcode > > I believe the US has s similar system to the UK, like Rally describes, but > their postcode system is based on numbers only, like 13745-1452 which is > inherently difficult to remember. The UK system is much better as it > utilized letters as well as numbers. This makes the codes shorter and easier > to remember. > > What I propose that we do - before any government body or commercial company > does it, is to create a near perfect postal code system for the Philippines. > This could be a separate project form OSM, but closely linked - so that OSM > will always be the best and most updated map with all postal codes included. > The reason why I want to do it this way is to ensure that the address, post > code and geo coding information remains free for all, unlike most othr > countries where only companies with a big enough budget can legally get hold > of the data. > > Rally has gone into great detail already, which is immensely useful > information. With my experience of db design in this specific area, I can > solve a couple of the problems he is facing: > He says people uses different ways to describe the same address. This is and > will always be a problem. The address db should therefore be centered around > points, not the address itself. A point can be either a single X-Y > coordinate or an area. Each point can have more than one address attached to > it. We can mark one address as "primary" or "official" so that looking up > any of the other address will give you the official address in the end. > So if "#245 Brgy Sta Cruz, Magalang, Pampanga" is the physical location as > "10 Rizal Rd, Puruk 1, Magalang, Pampanga" (it's not, I just made it up), > then so be it: add both address pointing to the same point. > Also, addresses change. This is a fact, and it's unavoidable, so the system > should be prepared for it. Having "valid_from" and "valid_to" dates on > addresses is a good start. This way, no historical data gets lost. In year > 2045, we can look back and see what the address was called in 2009. > And don't worry about database space. You shouldn't need to feel sorry for > the computers, they're there to serve us, and disk space gets cheaper every > day (about half price every 18 months), so I wouldn't worry too much about > that. > > I agree that the Lot/blk/phase thing is a stupid system that should be > abandoned immediately, however, failing that, having a system that allows > more than one address per physical location (point) goes a long way. When > someone actually comes up with a good name for the streets within a sub > division and numbering the houses within it, we can add them right away > while still supporting the legacy lot/blk/phase style. > If we assign postal codes with care - in other words if you have a > lot/blk/phase system, but you assign each postal code for one side of the > street in one block only, then you the lot number can act as house number, > as it will be unique. > Better yet, as the system will have a "primary" address, say you're a > courier like Rally and you're printing the list of places to go - the system > can look up the address (postal code + house number/name being unique) and > print the "real" or official address rather than the one given. So someone > writes: > Lot 12, blk 3, phase 2 > Diamond-gold-celebrity-life-buy-a-lot-here-and-be-happy-ever-after > Subdivision > Sometown, Someprovince > XX4X 1XX > It's translated to > 59, 5th Avenue > Diamond-etc subdivision > Sometown, Someprovince > XX4X 2XX > > I am prepared to design the database schema and direct the development of a > piece of software that can do all of this. I suggest that we create an > organization that holds the ownership of this software (which will only be > used on our web server anyway, and we don't want idiots copying our server > software to set up slightly modified alternative versions all over the > place) but release the data under an appropriate open license, similar to > OSM, available to everybody and easy to update, so it will be kept up to > date. > Exchange of data between OSM and this system would be obvious, leading to > all addresses - and postal codes - in the entire Philippines being available > on OSM - down to house number/name and their exact location. > > Hopefully, the post office and all courier companies, mail order companies, > direct marketing companies, etc. will see the value of this and adapt it as > well. > > Well, this was a lot of stuff in one go, but if anyone will raise their > hands to say they want to be part of this (and also add what they plan to > contribute with), then please let me know. I will then set up a task > management system where we can enter all our requirements and start moving > towards development of a system for it. > > Ronny > > maning sambale wrote: > > Hi, > > I have been discussing with rally about adding addresses to osm. We > all know it's messy here in the Philippines but rally outlined an even > more messy situation! > I hope we can pool our minds together to come-up with a sensible > addressing format for the Philippines. > > I've started a wikipage to collect some notes > http://wiki.openstreetmap.org/wiki/WikiProject_Philippines/Philippine_addressing > > I suggest we adopt the Karlsruhe Schema and adopt all possible tags > that fit the Philippines. > http://wiki.openstreetmap.org/wiki/Address > > here's the email from rally (background rally works in courier service > so he knows this stuff very well): > ============== > > > housenumber : number or alphanumeric (eg 12-C, 5-A), or > LOT/BLK/PHASE no. > housename: for addresses on Avenues (eg. ortigas ave ext or > manila east > roads) that has no street number, but building names only > > street: sometimes blank (for sitios with no streets; of for > subdivisions > with no streets - using Lot/Block/Phase) > > state -- not applicable > for NCR: > > postcode: zipcode > > Note: In the US, "zipcode" combined with > "house number" and "street name" > are enough info for a package to be delivered. > because their expanded zipcode makes the equivalent > barangay, sitio, town, > city, state becomes redundant info. > a specific expanded ZIP points to a small specific area in > the country. So, > once a messenger reached that zipcode area, all he has to > do is look for the > street and house number. sometimes it includes APT or SPC > (apartment or > Space number). no need to specify condo or name of > townhouse (as they > consume extra space in the database). > > The dream is for every filipino living in a specific area, > to know his/her > expanded zipcode. maybe, all he needs to do is dial his > landline on his > area, and the computer answers back with the correct > zipcode; or simply ask > the barangay tanod for their zipcode. then, there's no > need to memorize the > name of apartment, condo, building name, project > subdivision. that way, typo > error don't matter as long as the zipcode is correct. > according to wiki, > Taytay Rizal is 41square Km. yet the zip code is one > 4-digit number for the > entire town. Now what's the use of this zipcode to a > messenger? no use. > > my point is, after the tag "addr:street" we must > go to many other sub-levels > before going to city or town. > > some barangays in the philppines are just too big for one > expanded zipcode, > so we further divide it into sitio or big village or big > subdivision... > wait, it's not that easy to teach and implement. > > what complicates the address system is the inefficient use > of address fields > in paper forms: rural and city addresses need different > forms. > > NCR have "districts" and barangays but don't > have provinces, and sitios and > puroks are not that many. > > RURAL AREAS have provinces and barangays, but have lots of > sitios and > puroks, but practically no districts (except for > congressional districts > which is not used on Address system) - because > congressional district keeps > on changing very often. > > since there are lots of "streetname duplication" > in a particular town, rural > addresses need to specify subdivision name, or sitio or > purok name or > neighborhood association name; else the mails get lost. > There are no laws > governing the "naming of streets". it's a > free-for-all system for LGU to > name it to any names they like. > > if we use all fields (1) house number or Lot/Blk/Phase, (2) > Floor Number, > (3) building name or Apartment name, (4) streetname, (5) > sitio or purok, (6) > subdivision name (or project name), (7) barangay name, (8) > town or city > name, (9) province, (10) regions name, (11) country name, > (12) zip code... > gosh there's a lot of field that will be blank or be > filled up with a lot of > N/A's > because: > 1. some don't have House number, but Lot/Blk/Phase --- > How do you search Lot > Block Phase? Doesn't make sense and creates logistical > nightmare. One block > normally covers 3-4 streets (on all sides) - not > environment friendly - lots > of fuel spent searching. > 2. If Lot/Blk/Phase is used, then some don't use street > names (especially > for those subd given permit to sell even without naming > their streets), then > after a few years put names on the street and mess up the > address database. > 3. if they don't live in a building, then what's > the use of Floor Number and > building name field? (imagine the precious terabytes wasted > on the harddrive > trying to reserved space for these unused space on a > typical database) > 4. some buildings are condos, some office/residential (so > we use "building" > field to enter the condo name) when some would insist that > it should be on > the "Projects Name" --- same field reserved for > "Subdivision Name" so which- > which? what about a townhouse, since these are not > "buildings" but just a > "project". There are different impressions from > different people, so they > either put it on building (where they normally put condos), > while others > will put it on Subdivision (where they put the name of the > project). What > about warehouse factories with lots of building names > inside an industrial > subdivision? this is easy... > 5. some subdivision such as beverly hills subd (antipolo) > is considered "one > whole barangay" by antipolo city; so where do we put > it? in "barangay" or > "subdivision" field? or both? if both, then what > happened if we print it to > the mailing envelope using mailmerge... we get > "beverly hills subd, beverly > hills subd, antipolo city --- let's you think there > must be a typo somewhere > (because of redundancy). > and to think that half of beverly hills subd is also part > of Taytay Rizal > and is just under Barangay Dolores of Taytay. (there's > boundary dispute) > 6. some use sitio as another name for purok (in our town, > that is the case). > Sitio is the name, while that name has an assigned > "Purok #". Which one > should we put on the address field, the sitio name or the > purok number? so > stupid and confusing system, don't you think? you > can't have both if they > mean the same. but if you let people use either, your > database search gets > messed up. > 7. some Neighborhood association (which should be in the > "subdivision" field > is assigned a "purok" number). While a > "subdivision" can also be assigned to > a "barangay" field (see number 5). So who is the > final arbiter which goes > where, when filling up an address form? Answer, an LGU > official registry of > verified address. If you're address don't appear on > the system, then you > apply for one, verfied by the tax mapping if such address > is under what > taxdeclaration number, and connected to the Address > Registry Office if a > street number, house number, building number, apartment > number etc is > already assigned to it, or if it does exist. > 8. If it's a city within the province, then what do you > put in the province > field? blank? We normally dont put Antipolo City, Rizal > Province (because > antipolo has it's own congressional district) > 9. if the city is in Metro Manila, then it would appear > like Makati City, > Metro Manila. In which case, what in what field do we put > "Metro Manila", in > the province field? (Metro Manila is not a province). If we > use NCR instead > of Metro Manila, then I guess, we need to put another field > for Region, > because NCR (or Metro Manila) > http://www.wowphilippines.com.ph/explore_phil/explore_main.asp > Except for Metro Manila, Region was never a part of address > system. So why > put Metro Manila as an exception. It's a waste of > database field. gosh, we > need to teach all these stupid rules to all filipinos as > early as grade 3 > else they mess up the database, mess up our mail > deliveries, as well as > those ambulance and police rescue, and pizza deliveries. > 10. what about those vacant space in rural areas. No > problem in vacant > subdivision lots, because chances are, there's a > lot/block/phase number > assigned to it. In rural settings, rarely do an LGU assign > a SPACE number to > make it reacheable. Eg. what if you want a truckload of > panambak delivered > to it. Or you want to rent it for Perya or Bingo Tent > operations. In the US, > they lease these spaces to RV-truck houses. Mails and pizza > can actually be > delivered to these SPC (space). > > we have lots of these confusing address cases. so where do > we start? let's > build a matrix kaya? i'll supply you with actual stupid > addresses made by > some people that are giving us headaches and eats up our > earnings because of > multiple failed delivery attempts. > > let's crunch first the above puzzles above, and make a > fool-proof > idiot-proof efficient address system. :-) the resulting > databse field will > be the one used in data collection for our pilot project in > my area. > > > > -- cheers, maning ------------------------------------------------------ "Freedom is still the most radical idea of all" -N.Branden wiki: http://esambale.wikispaces.com/ blog: http://epsg4253.wordpress.com/ ------------------------------------------------------ _______________________________________________ talk-ph mailing list talk-ph@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ph