Final touches for the API

[fjarri]

Synced with https://github.com/nucypher/rust-umbral/pull/50

https://github.com/nucypher/nucypher/pull/2612 uses this branch in CI

• add VerifiedKeyFrag.from_verified_bytes() (for storage purposes)
• remove metadata support from reencrypt() (see https://github.com/nucypher/nucypher/issues/259)
• remove unused dependencies
• rename pk to delegating_pk in encrypt()
• rename decrypting_sk to delegating_sk in decrypt_original()
• rename decrypting_sk to receiving_sk in decrypt_reencrypted()
• expose serialized object sizes as HasSerializedSize.serialized_size(), and overhaul Serializable given the new assumption that all of our objects have constant sizes known in advance. The size is needed in nucypher (to set up splitters)
• in addition, split Serializable into Serializable and Deserializable
• replace GenericError with the standard ValueError

[cygnusv]

• rename pk to delegating_pk in encrypt()

Rather than delegating_pk, this should be a simple encrypting_pk. One of the core propositions of PRE is that delegation is completely orthogonal to encryption.

• rename decrypting_sk to delegating_sk in decrypt_original()
• rename decrypting_sk to receiving_sk in decrypt_reencrypted()

I don't get this distinction between delegating and receiving. In both cases it's "receiving"; "decrypting" also works.

[fjarri]

I don't get this distinction between delegating and receiving. In both cases it's "receiving"; "decrypting" also works.

Here's my rationale:

First, the encrypting/decrypting prefixes. I think they are redundant; if you are passing a key to encrypt/decrypt, these keys are by default used for encrypting/decrypting (while other usages still warrant a prefix, like verifying_key).

Now for the delegating/receiving prefixes. While it's true that one can use Umbral as a simple asymmetric encryption scheme (and we do use it that way to encrypt kfrags for Ursula), it is not the intended usage, and, in fact, an overkill. I wanted to emphasize that:

• during normal operation encrypt is supposed to take a key the delegator, and not the receiver, has control over (I was confused by this myself on several occasions)
• decrypt_original is more of a debugging tool than a part of a normal process, and takes the same key that was used in encrypt
• decrypt_reencrypted takes a different key compared to what was used in encrypt