As Lockridge pointed out, the errors and losses he was talking about are self-consistent, and, in the absence of any means other than the altimeter to know one's exact altitude, it is impossible to detect them. It is possible, however, to simulate them on a computer, and I did. (I sell the program CMARC, which I used for these analyses, so this is not only a Technicalities column but also a thinly veiled advertisement.) Mountains come in all shapes and sizes, of course, but I modeled a ridge of what seemed like a plausible shape, and also a pass. They are idealized forms, but once you get some distance away from the surface, the wind doesn't care about every bush and rock; it's the overall shape of things that counts.