In replacing Coord with SpherePoint (
DM-11162) I discovered a problem in the skymap package: skymap makes heavy use of Coord.offset, including for points at a pole. However, SpherePoint rejects such offsets (as well as negative offsets).
I believe the behavior of offset for a coord at the pole is unambiguous and unsurprising if the coord is constructed from longitude, latitude. However, I acknowledge that this is less so for a coord constructed from a 3-vector. Nonetheless, I propose to change the behavior of SpherePoint to match Coord and document the behavior at the pole.
I also propose to allow negative offsets, as the intent is unambiguous for those and Coord supports it.
Also make sure the sign convention for the orientation of offset is the same for SpherePoint as Coord and document it carefully.