Real objects are basically always either psuedo-radix because you have a good idea of the distribution and aren't limited to pairwise comparison, or psuedo-selection because the objects are big and heavy so you don't want to move them more than necessary. For people self-sorting psuedo-radix is definitely the way because they can self-parallelize easily.
I say psuedo because real objects have lots of properties both on the rearranging and comparison sides that are not quite the same as digital ones. For eg it's faster to compare numbers that have a large difference in magnitude, and it's easier to swap objects that are physically close to each other. So following a CS algorithm to the letter is usually slower than adapting a little.
I say psuedo because real objects have lots of properties both on the rearranging and comparison sides that are not quite the same as digital ones. For eg it's faster to compare numbers that have a large difference in magnitude, and it's easier to swap objects that are physically close to each other. So following a CS algorithm to the letter is usually slower than adapting a little.