[uf-discuss] geo - accuracy of coordinates

[Kevin Marks]

On Oct 2, 2006, at 4:11 PM, Michael MD wrote:
> Not really, if it's a large city...
>
>> Consider Birmingham, England, whose "centre" is far from being
>> equidistant to all points on its boundary - it's in "Ladywood" on this
>> map:
>>        http://www.birmingham.gov.uk/wards
>>
>> GeoRSS uses a "radius" element, which could be a uf class and would
>> specify in meters, kilometers, or miles (itself a discussion for
>> units). Or perhaps better would be a featuretypetag (again using
>> GeoRSS as a working case example) that can specify "city"
>>
>> Or the capacity to describe a polygon...
>
> In many cases the available data is just not accurate enough to be  
> able to
> accurately describe a radius or polygon.
> (if it were, it might be accurate enough for a street address too!)

Again, consider this URL:

http://flickr.com/map/?&tag=yankeestadium&fLat=40.828081&fLon= 
-73.920821&zl=7

if it were expressed as a radius of 10km, I think it would do for Brum  
as a whole.

> Even with only city-level accuracy, lat/long data can still be very  
> useful,
> but clearly I'd want to avoid having it end up being used to generate a
> local street map that gives someone incorrect directions to get to a  
> street
> address!
>
> The radius idea could possibly be used but the radius itself would be
> inaccurate.

That's fine - it's really an order-of-magnitude expression of detail.

>  I think the featuretype idea is probably better for this if
> there can be a standard for the actual feature type names.
> (but in GeoRSS it appears to be folksonomy-based so if data comes from
> multiple sources, an application might need to look up some kind of
> dictionary of commonly used terms just to work out that it's a "city")

Or, as in this example, it could be derived from the tag + geo  
expressed by several people.