GeoNames Forum

Hi David,

the file that marc posted was a temporary dump. It does not really make sense to import this file as it may already be out of date.

The best strategy is probably to use the same algorithm as marc to attach the PPLX to their respective PPL :

-> For each PPLX being imported, search for the most populated city (PPL, PPLC, ...) within 10 km. Attach the PPL to this city

OK thank you !!!

Hi,

in Adm1ASCII file, half of the US states have a leading "State of .." in their name" (which is not true in the other adm1 file).

For instance, (Adm1ASCII) :
US.CA State of California State of California 5332921

but (adm1)
US.CA California

This is not the case of New Jersey, for instance.

Is there some logic in that ? Do you want me to remove the trailing State of manually of the US states ?

regards,
Sami Dalouche

Hi,

Does anyone know what exactly all the Junctions (especially in the US) are ?
http://www.geonames.org/search.html?q=junction&country=US

Do they really qualify as 'PPL' ? For instance, if you're searching for Cities around Los Angeles, all these junctions are more polluting (street junction, taylor junction, ...) than giving any meaningful result.

So, if someone has an idea of what these junctions exactly are, it would be nice to explain, so that we can change their feature codes...

thanks,
Sami Dalouche

OK, thanks

will be easier next time

Hi,

http://www.geonames.org/3013131/paris-04-hotel-de-ville.html is marked as 'PPL', but is in fact a 'PPLX'.

I cannot change this myself since
'error while saving:
the name is locked and may only be updated by users with a user level >= 2. Your current user level is 1'

So, can anyone with a level of 2 change it, please ?

thanks,
Sami Dalouche

Hey,

well, in case they wouldn't accept removing the copyleft, wouldn't it make sense for geonames to change its license by adding a copyleft ?

When you think about it : a copyleft could protect your project and make sure nobody takes the geonames data and releases it as something non free.

Another possibility might be to provide dual licensing. For instance, there could be a 'light' version of geonames available under CC-BY, and the 'full' version available under CC-BY-SA.. That could probably cause confusion, but it's worth thinking about it, isn't it ?

Regards,
Sami

See : http://www.zillow.com/labs/NeighborhoodBoundaries.htm

Does it make sense for geonames to integrate the data ?

Sami

Hi David,

I also plan to extract the GIS stuff from my code (that uses Spring/hibernate/ bleeding edge stuff) and would be happy to contribute to the project as soon as I find some spare time...

Regards,
Sami Dalouche

Hi,

District data is sometimes present in geonames as "PPLX" features, which means Section of Populated Place. (See : http://www.geonames.org/export/codes.html ).
You can find some of these PPLX here
:
http://www.geonames.org/advanced-search.html?q=paris&country=FR&featureClass=P&continentCode=

The problem I see with the way this data is structured is that the PPLX are not strongly linked with their respective PPL*. So, you basically have to pre-process the data to get the relationships, and the methodology for doing this is not really clear... I think marc said once in the forums that he was basically doing GIS queries around (e.g. 10km) the PPLX, sort the PPL's by population, and link the PPLX to the most populated PPL.

This should work in most cases, since there are rarely 2 cities that both have PPLx and are distant by less than 10/20 km...

Cheers,
Sami

Oh, also :

in ADM1 ASCII file :

EH. EH. Oued Ed-Dahab-Lagouira 6547304

Hi,

I believe the wisest thing to do is to stick to the current ISO codes and let the ISO guys handle the political issues.. If ISO says it is an independant country, then I think geonames should blindly follow ... (BTW, I thought that the issue was about algeria vs maroco's western sahara, not maroco vs independant ;-p))

Anyways...
Just to tell you, today's dump (September, 16th) still contains the error :
MA..066 Aousserd Aousserd 6547297
MA..391 Oued-Ed-Dahab Oued-Ed-Dahab 6547298

Regards,
Sami Dalouche

Hi,

(I cannot change the lat/longs directly from the web interface, so .... )

What is the main source for Canada's Lat/Longs ? It looks like thelat/long for "Saint Lambert", near Montreal is incorrect, and I wonder if other canadian cities have incorrect data too...

http://www.geonames.org/6138599/saint-lambert.html advertizes 48.95019 / -79.46636 whereas http://fr.wikipedia.org/wiki/Saint-Lambert_%28Mont%C3%A9r%C3%A9gie%29 advertizes 45° 31’ 20’’ Nord 73° 30’ 37’’ Ouest .

http://maps.google.fr/maps?f=q&hl=fr&geocode=&q=montr%C3%A9al&ie=UTF8&ll=45.497684,-73.511581&spn=0.310435,0.6427&z=11&om=1 seems to correlate with Wikipedia...

Regards,
Sami Dalouche

Marc,

Creating a temporary RR code will be fine for now, I think.

What is your policy regarding the codes ? Are the Adm 1/2 codes the FIPS ones, or the ISO ones ? Or maybe a mix of both ? From your data schema, will you be able to tell, later, that the 'RR' region code isn't an official ISO code ?

Regards,
Sami

Hi,

In ADM1Ascii, there is a broken line :
TJ. Region of Republican Subordination Region of Republican Subordination 6452614

which should be corrected to :
TJ.RR Region of Republican Subordination Region of Republican Subordination 6452614

(http://www.statoids.com/utj.html , or http://en.wikipedia.org/wiki/Region_of_Republican_Subordination)

Additionally, something like the following line can be added in ADM1 File :
TJ.RR Region of Republican Subordination

Regards,
Sami Dalouche

- Concerning the KR region : I think the ISO code is fine. (will you also add it to the adm1 file ? or just add it to the adm1ascii file)
- Concerning the ASCII equivalent : a null value is fine. I don't think people should use this field anyways. (lucene) text analyzers & stemmers are a better way to prepare the data for full text search.

Regards,
Sami Dalouche

A few corrections in adminASCII file: (you told me that the ADMs cannot be changed from the UI, right ?)

GB.A4 Bath and North East Somerset Bath and North East Somerset 6457408

And some additional non ASCII names :

LV.00 Jurmala Jurmala 459202

MA.00 El Aayoune El Aayoune 2543878

MQ.00 Département de Martinique Department of Martinique 3570311

MR.00 Disctrict de Nouakchott Nouakchott District de 2377449
MV.00 Maale Maale 1337624

And I don't know what to do about this one, since there is no equivalent in ADM1.txt :
TJ. Region of Republican Subordination Region of Republican Subordination 6452615

Should we remove it ?

Regards,
Sami Dalouche

Hi Marc,

I do think addresses and other POIs such as filling stations, supermarkets, etc are interesting to have in the daily dump.

So, if you have the time as well as the required authorization to include them, this data is welcome !

Regards,
Sami Dalouche

Hi,

a few countries did not have any currency :
-----------------------------------------------------
Antarctica
Bouvet Island (Bouvetoya)
South Georgia and the South Sandwich Islands
Heard and McDonald Islands
British Indian Ocean Territory (Chagos Archipelago)
United States Minor Outlying Islands

I have updated the countries file to include the countries of these countries (except antarctica, that doesn't seem to have any)
http://www.sirika.com/data/geonames/geonamesCountries.20070526.txt

The currencies have been set according to :
http://www.brokersmatrix.com/index.php?pag=resources&act=currencysymbols

Changelog:
IO IOT 086 IO British Indian Ocean Territory (Chagos Archipelago) 60 0 AS .io USD Dollar +246 en-IO 0

UM UMI 581 - United States Minor Outlying Islands - 0 0 OC .um USD Dollar en-UM 0

BV BVT 074 BV Bouvet Island (Bouvetoya) 0 AN .bv NOK Krone 0

GS SGS 239 SX South Georgia and the South Sandwich Islands Grytviken 3,903 100 AN .gs GBP Pound 3474415

HM HMD 334 HM Heard and McDonald Islands 412 0 AN .hm AUD Dollar 0

Regards,
Sami Dalouche

Actually, I tried feeding all the missing names to postgres, and took the minimum levensthein distance between the missing name and all ADM1s of the given country.. The result is really bad, and completly unusable...

Maybe you have a better heuristic ?

Regards,
Sami Dalouche