[Haskell-cafe] MD5?

[Derek Elkins]

On Sat, 2007-11-17 at 16:40 +0000, Andrew Coppin wrote:
> Thomas DuBuisson wrote:
> >> BTW, while I'm here... I sat down and wrote my own MD5 implementation.
> >>     
> >
> > How is the performance on this new MD5 routine?
> 
> Ask me *after* I modify it to give the correct answers. ;-)
> 
> Interesting question: How do you determine when an implementation of 
> something as complex as MD5 is actually "correct"? I might get it so it 
> passes all the tests I've tried, but there's some obscure edge case that 
> makes it fail. How would you know that? Hmm, in fact... how do I know 
> the implementation(s) in checking my program *against* are correct?

It's a small enough program that it should not be that hard to prove
that it does the same thing as the reference implementation (although,
that one may be broken...)