[noise] snow: a rust implementation based on screech

[Jake McGinty]

> On May 30, 2017, at 8:56 AM, Trevor Perrin <trevp at trevp.net> wrote:
> 
> Cool, I assume that's 250 MB/sec?

It’s genuinely showing ~2500MB/s assuming ring’s benchmark tests are also accurate (results matching the similar AES-GCM TLS tests done in https://github.com/briansmith/crypto-bench when run on my laptop). Running on another server just now I see ~1500MB/s.

But if that sounds unrealistic (I’m also skeptical of that performance in production settings), feel free to scour https://github.com/mcginty/snow/blob/master/benches/benches.rs <https://github.com/mcginty/snow/blob/master/benches/benches.rs> and let me know what’s wrong. I haven’t checked the compiled output to see if anything’s getting optimized away, etc.

> Does *ring* not provide BLAKE2 or X25519, currently?

*ring* unfortunately doesn’t provide BLAKE2, and while it supports X25519 it doesn’t allow you to provide your own private key yet. If I have time this week I was considering submitting a PR to get that in though.

Once I also add in feature-gated support for libsodium via a library like sodiumoxide, though, we will have a much better X25519 implementation available quickly.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://moderncrypto.org/mail-archive/noise/attachments/20170530/77d0e854/attachment.html>