Remove padding using new encryption API #246
Conversation
The signing and verification is now implemented using the generic Signature and *Signer / *Verifier traits. Drop old API proprietary to the RSA crate. Signed-off-by: Dmitry Baryshkov <[email protected]>
Signed-off-by: Dmitry Baryshkov <[email protected]>
Add traits following the signature design for encryption and decryption. Signed-off-by: Dmitry Baryshkov <[email protected]>
Drop encryption/decryption from the RsaPublicKey/RsaPrivateKey structs, add new EncryptingKey and DecryptingKey structs implementing Encryptor / Decryptor traits. Signed-off-by: Dmitry Baryshkov <[email protected]>
The crate doesn't use anymore the PaddingScheme. Drop corresponding API. Signed-off-by: Dmitry Baryshkov <[email protected]>
The function mgf_xor1() was replaced with mgf_xor1_digest(). Drop the legacy function. Signed-off-by: Dmitry Baryshkov <[email protected]>
|
This PR incorporates #245 instead of replacing it to let us discuss these two PRs separately. |
| //! ``` | ||
| //! use rsa::{PublicKey, RsaPrivateKey, RsaPublicKey, PaddingScheme}; | ||
| //! use rsa::{RsaPrivateKey, RsaPublicKey}; | ||
| //! use rsa::pkcs1v15::{DecryptingKey, EncryptingKey}; |
There was a problem hiding this comment.
EncryptingKey reads quite odd, I would have expected it to be called EncryptionKey
|
It's a neat API, but from a convenience perspective I am not a fan honestly. Having to create new types for single function calls just does not make for a simple to understand API, especially given that the key is not actually changed, it just stores some metadata. So I honestly think we should not remove the legacy API, or at least provide a better convenience API on top of this structure. |
|
I do like this approach overall. I worry there's going to be a lot of complaints and confusion for anyone attempting to upgrade who just wants to be able to call an inherent method on a key type. If we do go with this I think it needs some documentation in the changelog or elsewhere about how to upgrade from the previous API. |
|
Re: this vs #244, I think both would be interesting? I think there are plenty of use cases for #244 where a given key might be used for any application with any padding mode, especially KMS/TPM-style applications where a given client might ask for a key operation involving any of them (i.e. both signing and encryption, depending on the client's whims). However, if you know ahead of time what the key's usage will be, the API in this PR is great. |
|
@lumag if you can rebase, the encryptor/decryptor types seem interesting |
|
Closing in favour of #259 |
3 participants
This is a reimplementation of RSA encryption/decryption API on top of #245 .
cc @tarcieri @newpavlov @dignifiedquire