The idea is simply to write comparative tests into the code; you have the "proven but unsupported" production method and then, at its side, a very "safe" method that performs exactly the same calculation, on a small subset of the data.
Why wouldn't you use the safe method every time? It seems to me that if you have a method that works and is safe, why would you not use that? Maybe I'm missing the point completely. :hehe:
Go Badgers... (NCAA)