I'm examining the two geospatial options for IBM's DB2 database. In particular I am interested in the accuracy between these two options and whether it's worth it (the more accurate representation isn't free).
The thing I am most concerned about is someone doing a query (e.g. looking for a point or other geometry within a polygon) and the database incorrectly returning an empty resultset. My use case is a catalogue indexed by several means (temporal, geospatial and XML) and entries can be anywhere on the globe.
IBM states (PDF link) that:
Spatial Extender treats the Earth as a flat map. It uses planimetric (flat-plane) geometry, which means that it approximates the round surface of the Earth by projecting it onto a flat plane. This projection causes distortions, which can vary across the extent of the data, but the distortions generally increase toward the edges of the projected region…Spatial Extender is best used for local and regional data sets that are well represented in projected coordinates, and for applications where location accuracy is not important.
The other option is:
Geodetic Extender treats the Earth as a globe. It uses a latitude and longitude coordinate system on an ellipsoidal Earth model. Geometric operations are precise, regardless of location…Geodetic Extender is best used for global data sets and applications that cover large areas on the Earth, where a single map projection cannot provide the accuracy required by the application.
If I was using something like WGS84 in the Spatial Extender geospatial implementation, can anyone give me a ballpark figure of the worst-case accuracy errors I might encounter when compared to using the Geodetic Extender? Is there a formula or process for working it out for anywhere on the globe for a given map projection?