Joe Birr-Pixton

> Now that #1963 is close to merge, how would we like to handle bogo testing w/ `*ring*` as the crypto provider w.r.t the ECH tests? Can we do more...

Have pushed a new revision of this: - addressing the above review comments - added quite a bit of commentary at the crate level - made explicit ciphertext/share types in...

> so I'd expect the crate to be named rustls-post-quantum instead? Renamed.

the intention is that `CryptoProvider` is the only mandatory trait that needs implementing -- everything else is optional. there's no requirement that one `CryptoProvider` provides everything -- eg, it's absolutely...

> wanting to export the CryptoProvider API in FFI So I don't think we should provide FFI APIs that allow custom `CryptoProvider`s to be written. I think for most existing...

I suggest when transferring the tags after this merges, we rename them `(subdir)/(semver)`, eg, `pki-types/0.2.1`?

And for the workflow yml, I suggest we name that by the subdirectory name too? If we keep a separate pki-types workflow, should we restrict it to running for changes...

> I think we can get away without separate workflows for now, especially since pki-types is very fast to build and we can relegate slower parts to the daily tests...

I think I'm still a bit confused about the interaction between the ci.yml being added here from pki-types, and the preexisting build.yml?

See also https://github.com/rustls/rustls/commit/22f083e404c769efd4536097c5c7933a7f2016bd and https://github.com/rustls/rustls/commit/0cca8222ed25d42c3af7f1c1d907428f8189d7f5 which is attacking the same problem, albeit quite a while ago now. I won't rebase that branch (I was mainly interested in how much performance...