Assuming I'm building a 3D solution with the "world" being mapped to global geo-data supplied by OSM / Cloudmade... think http://www.webglearth.com but with completely different goals and use-cases. In fact we won't have a round sphere but only work off a "flat" (+ elevation later) surface/ground plane at all times.
Now assume further that we want to texture the "visible" fragments of the "ground" plane with OSM and/or Bing/MapQuest/Google aerial map tiles, the level of detail / zoom level to be chosen based on the "camera's" / viewer's "altitude" / distance-to-ground -- say in meters.
Knowing that 1. we only use OSM-equivalent tiles using exactly the same spherical mercator projection (which, from what I gather, is used by all the major map players ie. OSM, Google, Bing) and hence 2. always equivalent scale-distortions depending on latitude: how can I decide which zoom level to switch to when viewer altitude changes?
Of course the boundary cases are easy: zoom level 18 is to be used on-or-close-to-the-ground and zoom level 0 from "outer space". But the intermediate cases are a mystery to me when it comes to figuring out "from which altitude they had been taken if they were simple photos", so to speak.
I saw this page http://wiki.openstreetmap.org/wiki/Zoom_levels and some related pages but nowhere did I find an approximation on how to "map" a fictive altitude to a zoom level.