A comment by Edwin T. Jaynes, from "On the Rationale of Maximum-Entropy Methods" (Proc. IEEE v.70, n.9, Sep 1982, p.947):
... for 200 years applications of probability theory have been plagued by the seeming impossibility of communicating to another person exactly what problem is being solved. Dating back at least to Laplace, almost every writer on probability theory has had the experience of giving the correct solution to a problem, only to have it rejected because it is not the solution to some different problem. And indeed, an old joke among mathematicians runs, "I have found a beautiful solution, but I have not yet found the problem."
(cf Programming vs. Thinking (2013-01-18), ...) - ^z - 2013-02-09