If you do this, you still need to do a one-time correction for the length of a degree longitude at a given latitude, or you get weird behavior even at small dis

If you do this, you still need to do a one-time correction for the length of a degree longitude at a given latitude, or you get weird behavior even at small distances (and it gets weirder the closer you are to a pole). We learned this the hard way when our "k-nearest places" query was returning point clouds which looked like ovals!