Encryption features > Discussion on SignedFooterEncrypt

[argmarco-tkd]

At this point, I am not sure whether we need to implement this method or not.

Originally, it was included because of this code:
https://github.com/apache/arrow/blob/71894728810fa2c191a95315574fa5ea4546566f/cpp/src/parquet/metadata.cc#L2094

where I thought we needed to replace the hardwired version of AES_GCM_V1 with the actual encryption algorithm used.

Turns out, this code is only executed when the file has a plaintext footer. In that case, yes a hard-wired AES_GCM_V1 encryptor is used to sign the footer, but that is not related to the rest of the encryption.

(There is a separate issue, and that is that this hard-wired value is later mis-interpreted as the file level encryption algorithm, but that is being tracked in issue #95 )

So we need to decide: should a plaintext footer always be encrypted using AES_GCM_V1? Or do we want to use whatever the file-level encryption algorithm is? My gut feeling is to leave that unchanged, as it doesn't affect our code.

Thoughts?

----- Original issue description ----
@sofia-tekdatum : Explain why this is not needed

Resolve the question: should the SignedFooterEncrypt() method be part of EncryptorInterface?

The method is used for footer verification inside parquet/metadata.cc, via direct instantiation of AesEncryptor).

Link to discussion during PR review: #68 (comment)

[sofia-tekdatum]

We will not implement this method, we won't change Arrow's functionality, we'll let it use the default AES_GCM and we'll leave the method in the interface, and throw an unsupported exception in our code.

[argmarco-tkd]

and throw an unsupported exception in our code.

@sofia-tekdatum - do we have an Issue for tracking that work (exception throwing)? is this going to be that issue?