I am not the author of this tag but I have thought about it.
If the intention is to determine if an itinerary is passable or to warn drivers life about dangerous turns, rather complicated math has to be done for each and every bend and it's impossible to do it instead of using turning_radius because of the usual imprecision of OSM. It's also overkill for a smartphone GPS to do it in real time. I, and probably all GPS writers, would refrain from letting drivers believe in safety with that math. turning_radius is supposed to be calculated precisely and only in dangerous places.
I agree that a discussion is needed, but for other reasons. One is that if should be specified where the tag is put, probably on the node with smallest radius. Someone who has looked at the existing turning_radius tags is not only convinced that their radius cannot "be calculated from the OSM vector data" but is also wondering which turn it's about when placed right in the middle of a junction. Also, if using turning_radius were used on roads, shouldn't the sections of road that have been surveyed be indicated to know where the radius of untagged bends is above which value? --Papou (talk) 12:40, 31 August 2013 (UTC)
Axles follow different paths
For tight bends and long vehicles, isn't there a noticeable difference at the turning radii of different axles? Where the front wheels need a radius of, say, 12 meters, the back wheels follow a different path? On "normal" roads it's not a real concern, there's excess width on the lanes and on the shoulder, but in the forest/mountains it might be needed more? Alv (talk) 14:51, 31 August 2013 (UTC)