The proposal should include prior art, since there has been many, particularly for Rust, e.g., Joshua's RFC implemented in Fuchsia and based on FromBits/IntoBits pre-RFC implemented in the simd crates, similar traits in the TTF library, the interprocess crates, safecast, etc. This all amount to many years of experience trying to solve this problem in Rust in practice. For example, the Compatible<T>-trait proposals look like a much better overall direction for safe transmutes than this proposal to me along all axes I can think of, so it would be interesting for this proposal to include rationale about why this should be done instead. I'm probably missing some obvious tradeoffs, but it would be good to have them in writing.
1 Like