Limits on Inputs to LabeledExtract and LabeledExpand about draft-irtf-cfrg-hpke

[rozbb]

A side effect of this would be that the Psk and PskBundle (containing a PSK and PSK ID) types in the Rust implementation would not need to allocate and also not need cumbersome lifetime specifiers, since they could just maintain a fixed-size buffer of their contents.

[chris-wood]

What would be "reasonably small" here? TLS allows 16-bit PSK identities and exported contexts, for example.

[rozbb]

If pskID were 2 bytes (or an integer that we explicitly encode) and everything else were at most size Nh, that would be sufficient

[blipp]

Adding the following just so that we remember that in case we introduce limits: If in the end the PSK's size is limited to smth strictly smaller than Nb (block size of underlying hash function), we could get rid of the hashing step for the PSK.

[rozbb]

Update: this change may not be necessary. It shouldn't be hard to get HKDF-Expand and HKDF-Extract to work this way without allocating.

Waiting on RustCrypto/KDFs#34

[rozbb]

KDF PR got merged. This is no longer an issue.