[OSM-talk] Wood & Park mapnik carto anomaly?

[Paul Norman]

On 9/17/2014 7:33 AM, Dave F. wrote:
> I wouldn't describe size based ordering as 'well defined'.
The polygon table has no column or combination of columns which is 
guaranteed to be unique. We only need to have a well defined ordering 
within a meta-tile that is consistent across boundaries, not globally, 
and way_area works well for that. Thanks to precision limits in the data 
and conversions to Web Mercator two apparently identical area polygons 
will probably slightly differ in the exact calculated area, and two 
polygons identical in WGS84 area may not be identical in Web Mercator 
area. If two polygons are exactly the same area they probably are the 
same geometry, which is a special case.

A sort order that is well defined in nearly all cases* would be both way 
area and osm_id, but the additional ordering term would probably 
negatively impact the speed and memory of the sort.

* Again, there are certain exceptions in theory.