[OSM-talk] US City Import?

[Beej Jorgensen]

Robert (Jamie) Munro wrote:
> Beej Jorgensen wrote:
>> 3. Some cities are in the OSM database already.  Data duplication,  blah 
>> blah.
> 
> Can you try and look for them by comparing names with nearby names in
> Planet.osm, or using the zappy api?

I could.  All that would be left is to decide which data was "right". :)

Getting all the cities and towns in the US with the xapi is no problem, 
so I could read that in to not upload duplicates.

A few quick queries show that there are currently about 1000 cities, 
towns, villages, and hamlets in total in the US.

> I'd like to see the date alongside all the census tags. Then we could
> know how up to date the values are, and after future censuses we could
> keep historical data.

Something like:

	<tag k="population" v="3040;YYYY-MM-DD" />

?

This almost sounds like something that should be standardized, if we go 
this route.  The census data files I have go back a few years, so I 
could import the lot of it.

A question I have here is: is historical census data something that 
should go in the OSM database?  It can easily be crossreferenced by 
interested parties later from external sources.  Furthermore, wouldn't 
future population changes be automatically saved in the node history?

> If you make the files, someone can probably load it in to the Tiger bulk
> uploader that runs on the "dev" server.

This probably won't be necessary, but I could do it.  I estimate it 
would take a little less than a day to do the whole US from here, or 
about 2-3 hours if I only did the major cities.

Another question... I use the stock definitions of the city sizes, 
currently:

hamlet < 100  (I made this up--page says "a few" houses)
village < 10000
town < 100000
city >= 100000

This means that nothing in Montana, for instance, will be a "city". 
Should the city sizes be uniform over the whole US, or should I scale 
them in each state?  I'm for uniform, since things might look really 
weird on a state boundary.

-Beej