[Geojson] circles (was: Toward consensus on proposals)

[Arlo Belshee]

> (There seems to be a common belief in the GIS world that this is okay; I would argue very strongly that it is not.)

>From the perspective of GeoJson, I agree. Self-describing payloads are needed.

However, this is the norm within the GIS community (after all, all the big companies already have such a table, so why would they need the data in the payload?). And, honestly, a tremendous number off apps only use 1 coordinate system, so they can just hard-code in knowledge of what the distance member means.

No, it's not a good reason to go with the coordinate-system-dependent route. However, this is: what happens when someone reports circle radii in km, but computes distance (as simple features algorithm) in Indian Feet (the standard measure for some coordinate system out there)? That makes things difficult for the client, too.

Yes, that's a poorly-implemented server. They really should use the same metric for distances everywhere. But it's fairly easy to fall into if your radius is not explicitly bound to the same metric as your distance.

Arlo