Implementing crypto packages is a real pain.
I wrote an implementation of DES in Java in 1996. Absolutely not interesting for me.
 
For this kind of stuff, I think writing an oCaml wrapper on top of C library is better.
 
 
>
More seriously, perhaps, a well-known source of flaws is complexity.
Expressing algorithms in Ocaml is often (much) less complex than
in C...
 
-- That's probably true for public key cryptography. For symmetric key crypto and hash function, the algorithms are not elegant.
 
In companies, there is always the dilemma: make vs buy.
I think the oCaml community should think about "re-implement vs wrap". I don't think the real value of oCaml programming sits at the level of the OS or the low-level libraries.
 
Arnaud