Use crypton instead of cryptonite

[mpilgrem]

See yesodweb/wai#931. hpack-0.35.3 has moved to crypton - see sol/hpack@26ec126.

[hasufell]

crypton has never been audited either and there's no evidence that GHC/Haskell is not particularly prone to side channel attacks.

[mpilgrem]

@hasufell, I recognise your warnings from your contributions to this discussion https://discourse.haskell.org/t/a-new-future-for-cryptography-in-haskell/3888/5 and the Reddit thread on the fork: https://www.reddit.com/r/haskell/comments/14245q8/crypton_is_forked_from_cryptonite_with_the/.

In the case of Pantry and Stack and cryptonite, the direct use is only:

• SHA256 hashing (Pantry)
• MD5, SHA1 and SHA256 hashing (Stack)

I am assuming that if those particular functions did not do what they purport to do, somebody would notice sooner rather than later.

[cdornan]

The point is that crypton is being actively maintained, it is a better choice than cryptonite.

I get the sense that the crypton maintainer would like to address the underlying issues, but the first step is to execute the fork, no?

Thanks @mpilgrem for your incredible work — I feel we are all in your debt. (And too @hasufell.)

[hasufell]

Well true... but crypton is also a dependency of tls, which stack uses indirectly. But that is an old discussion.

[cdornan]

Indeed — I think tls was the main motivation for creating crypton.

[mpilgrem]

I started to implement this but got stuck with cryptonite-conduit. Raised kazu-yamamoto/crypton#13.