Last year Vaughan Pratt (of automata theory and Knuth-Morris-Pratt algorithm) asked on a hardcore theoretical physics group blog how does a light mill (device allegedly proving corpuscular nature of light, http://en.wikipedia.org/wiki/Light_mill) really work. To my knowledge this question after 150 years still awaits consensus over a satisfactory anwser (while having far too plenty seemingly convincing ones).

http://golem.ph.utexas.edu/category/2008/07/light_mills.html

A toy simulation:

http://www.gregegan.net/SCIENCE/LightMill/LightMill.html