feat: Make crypto implementation pluggable.

[ShogunPanda]

Hello.
This PR extracts all low level crypto interface in a specific file and makes the crypto interface pluggable in Noise.
Hope this helps!

[CLAassistant]

All committers have signed the CLA.

[mpetrunic]

Tnx for your contribution. Not a bad idea, will review in more detail next week

[mpetrunic]

lgtm, you should probably put latest master as we are converting to esm

[codecov-commenter]

Codecov Report

Merging #133 (590facf) into master (75116be) will decrease coverage by 2.67%.
The diff coverage is 92.81%.

@@            Coverage Diff             @@
##           master     #133      +/-   ##
==========================================
- Coverage   88.59%   85.92%   -2.68%     
==========================================
  Files          16       18       +2     
  Lines        1841     1918      +77     
  Branches      248      242       -6     
==========================================
+ Hits         1631     1648      +17     
- Misses        210      270      +60     

Impacted Files	Coverage Δ
src/noise.ts	73.96% <60.00%> (-2.56%)	⬇️
src/crypto/streaming.ts	97.67% <97.67%> (ø)
src/crypto.ts	100.00% <100.00%> (+2.32%)	⬆️
src/crypto/stablelib.ts	100.00% <100.00%> (ø)
src/handshake-ik.ts	61.39% <100.00%> (-30.92%)	⬇️
src/handshake-xx-fallback.ts	93.10% <100.00%> (+0.16%)	⬆️
src/handshake-xx.ts	97.15% <100.00%> (+0.03%)	⬆️
src/handshakes/abstract-handshake.ts	92.11% <100.00%> (+0.03%)	⬆️
src/handshakes/ik.ts	89.03% <100.00%> (ø)
src/handshakes/xx.ts	90.21% <100.00%> (ø)
... and 4 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 75116be...590facf. Read the comment docs.

[ShogunPanda]

@mpetrunic Rebased and updated!

[dapplion]

Looks good to me, nice abstraction! Are there any performance tests to ensure equal throughput? Results should be the same, but just to be sure

[wemeetagain]

LGTM from an organizational perspective

[mpetrunic]

Looks good to me, nice abstraction! Are there any performance tests to ensure equal throughput? Results should be the same, but just to be sure

master
$ node benchmarks/benchmark.js
Initializing handshake benchmark
Init complete, running benchmark
handshake x 40.53 ops/sec ±17.26% (72 runs sampled)
Done in 6.54s.

PR#133
$ node benchmarks/benchmark.js
Initializing handshake benchmark
Init complete, running benchmark
handshake x 41.62 ops/sec ±15.59% (72 runs sampled)
Done in 6.39s.

It varies on each run but it's identical

[mpetrunic]

@dapplion This potentially allows you in lodestar to use some native library that might be faster than stablelib (purejs).

[dapplion]

@dapplion This potentially allows you in lodestar to use some native library that might be faster than stablelib (purejs).

yeah looking forward to testing multiple impl

[ShogunPanda]

If you want, I have sample implementation with sodium-native in https://github.com/web3-storage/ipfs-elastic-provider-bitswap-peer/blob/main/src/noise-crypto.js

[dapplion]

If you want, I have sample implementation with sodium-native in https://github.com/web3-storage/ipfs-elastic-provider-bitswap-peer/blob/main/src/noise-crypto.js

Let's goo! 🔥 If you wanna contribute bringing this live in Lodestar I may airdrop you some $LODE tokens in the future

[MarcoPolo]

I've implemented this interface with WASM using Rust and the snow crate: https://github.com/MarcoPolo/js-libp2p-noise-wasm

Here's some benchmarks:

Throughput:

❮ node dist/benchmark/src/index.js
TCP+mplex+noise
Download throughput is (mbits/s) 369n
Upload throughput is (mbits/s) 272n
TCP+yamux+noise
Download throughput is (mbits/s) 193n
Upload throughput is (mbits/s) 195n
TCP+mplex+wasm_noise
Download throughput is (mbits/s) 1047n
Upload throughput is (mbits/s) 1055n
TCP+yamux+wasm_noise
Download throughput is (mbits/s) 869n
Upload throughput is (mbits/s) 867n

Handshakes per second:

❮ node benchmarks/benchmark.js
Initializing handshake benchmark
Init complete, running benchmark
handshake x 632 ops/sec ±0.58% (89 runs sampled)