Discussion: should isochrones be computed differently?

[sfendrich]

Problem statement

A discussion started in #1708 revealed that there might be a fundamental issue with how isochrones are computed in ORS.

The issue is that isochrones are computed from geo-coordinates. At the scale of a single isochrone, this is probably not relevant. But when comparing isochrones at different latitudes, the length differences of a longitude degree might affect the shape and size of the isochrone's geometry.

Expected outcome of this issue

Either (1) an explanation of why it is fine as it is, which should be placed as a comment at an appropriate place in the code or (2) a plan of what should be changed.

Questions to be answered in this discussion:

• Imagine an artificial town placed near the equator or near a pole. Would isochrones have the same shape in both places?
• Which parts of the isochrone computation are affected by such issues?
• At which scale is the approximation to take the euclidean norm of non-euclidean vectors inappropriate and are isochrones small enough to be unaffected?