Alternative Floe implementation in Rust

[poljar]

Hi there,

I started exploring FLOE and its Rust implementation last week, which led to a small PR: #9.

While digging deeper, I noticed from the discussion in #8 that there are some unresolved issues with the current implementation. As a result, I rewrote the Rust version: https://github.com/poljar/floe.

So far, I haven’t implemented the public streaming/online functions, only the semi-public (random access) functions are available. Even so, this new implementation offers several advantages:

1. no_std compatible.
2. Allows the user to choose the RNG implementation.
3. Uses generics consistent with most crates in the RustCrypto ecosystem, making it easy to swap in any AEAD/KDF implementation. For example, see the FLOE-GCM implementation.
4. Uses the zerocopy crate to parse and represent FLOE types (segments, headers, parameters) as higher-level objects.

There are some trade-offs as well:

1. More dependencies (zerocopy, thiserror).
2. Generics can be harder to understand.
3. The implementation is still incomplete.

With that said, I wanted to ask if you’d consider adopting this rewrite as the reference Rust implementation instead of the one being developed in #8.

[sfc-gh-grubin]

This is a good question and one we haven't considered. This time I really do mean it when I say "It will take some time to get back to you on this."

But thank you so much for your interest in this! We really do appreciate it and do want FLOE to be useful and a part of the larger cryptographic community.

[poljar]

No worries, it'll take some time to implement the rest of FLOE anyways.