[wp-hackers] Re: Place, Geotagging

[Steven Clift]

This informative reply is from Matt at MySociety:


Date: Thu, 06 Nov 2008 11:15:59 +0000
From: Matthew Somerville <matthew at mysociety.org>


On the wp-hackers mailing list, which I stumbled across on nabble, Mike 
Schinkel wrote:

> Steven Clift wrote:
>> One more link for the mix is the place name database that mySociety
>> uses for: http://gaze.mysociety.org/
>
> Very interesting. Glad to know about this. OTOH I don't see it working
> for this because of a few issues, in order of concern:
>
> 1.) It's a query mechanism that doesn't provide identity for places,
> thus "Atlanta, GA" returns 10 results instead of one (query UI at
> http://highearthorbit.com/projects/geocode/geocode.html)

Yes, it was never designed to be anything other.

> 2.) It doesn't publish any kind of service guarantee nor does it
> appear there is a good plan for for scability in place.

:-)

> 3.) Not sure what their policy/plan is for update of this data

Frankly, I think we should start using www.geonames.org and think you 
should probably look there too, they have many more data sources and 
presumably update often.

> 4.) It returns CVS instead of JSON or XML.

IMO, CSV is better than either of those two for this sort of data.

> 5.) It isn't really RESTful (a nit in this case, but something I
> notice.)

Not really bothered about that either ;)

>> I can't find the link, but I was once told by the developers that
>> they adapted a public domain U.S. military database of 4 million
>> place names around the world. Perhaps they now do something different
>> for various countries.
>
> It's here: http://geonames.usgs.gov/

To be precise, Gaze uses the geonet name server for non-US, and the GNIS 
for US. Plus other datasets to provide the population density functions etc.

> This database is very useful for other contexts, but it still doesn't
> get us to a unique placename identity. There currently is only one
> source for a unique placename AFAIK and that is Wikipedia.

No, there are others. For example, Flickr - 
http://www.flickr.com/services/api/flickr.places.getInfo.html - returns 
unique WOE (Where On Earth) IDs for places, which are defined by Yahoo! 
here: http://developer.yahoo.com/geo/

> To be
> clear,  Wikipedia can only guarantee uniqueness for an instant in time
> because they can change the URLs, although they rarely do.

Changing the URL wouldn't matter if it still followed, but being able to 
have any page deleted at any time might be more worrying (is this tiny 
village notable etc.).

>> What it does with a site like http://pledgebank.com in the U.S.
>> version is that when I search a city "Winona" it asks which one
>> specifically: http://www.pledgebank.com/search?q=winona
>
> Nice. I guess with sites that depend on it the web service must stay
> up...

Well, we do our best, but we're a tiny UK charity, so can't and haven't 
promised anything :-)

ATB,
Matthew, mySociety