Thanks for the suggestion. Using hashing is indeed good for checking data integrity. As you said, if a file is corrupt, one may hope to find a good copy elsewhere.

I want to achieve more than that. Using ECC, one can hope to be able to handle corruptions and so, a backup may still be good even if several bytes/blocks/sectors did go bad.

Storage is very cheap today. This means that increasing the size with redundant ECC information does not cost much. Another way to look at is that since storage _is_ cheap, one shouldn't count on it too much and devise a way to still maintain its reliability.