curve25519 wrapper

[xueyucoder]

python wrapper of D.J.Bernstein's curve25519 lib.

[xueyumusic]

python wrapper on DJB's curve25519 lib

[xueyumusic]

1. I *temporarily* put all of the sources in a dir , compile them into curve25519.a when using "python setup.py *", and use the new generated lib directly.

2. The python wrapper file is pycryptopp/publickey/curve25519module.cpp

3. The test file is pycryptopp/test/test_curve25591.py, which uses the data from DJB's curve25519 sources.

4. I use a shell script (create.sh) to create curve25519.a lib. The script will be executed when using "python setup.py build", however, when using "python setup.py test", it is also executed , which is not necessary.

5. This patch need be reviewed and modified. It is just a could-be-run version..

[warner]

Also see ​https://github.com/agl/curve25519-donna , which includes a python binding (that I wrote). The PublicKey/PrivateKey objects might make a nice API, or maybe there's some Crypto++ pattern that we should stick to. It's worth remembering that Curve25519 only offers DH key agreement (the output is a 256-bit shared random key). You can build public-key encryption on top of that, though (by creating a random keypair for each encryption, and including the public half along with the encrypted message, then discarding the private half).

[zooko]

pycryptopp now includes Ed25519 signatures, but not yet public key encryption. I think I'd like to go ahead and incorporate curve25519 public key encryption eventually, but not soon.

[davidsarah]

The right way to construct a PK encryption scheme from the curve25519 key exchange is probably to use ​DHIES/ECIES.

[warner]

Reassigning to zooko, because the unusual comma-joined "zooko, xueyu" owner was confusing some tools I'm writing that try to identify all real users (by looking at the owners of tickets, among other things).

[daira]

Why have this in pycryptopp rather than using one of the existing libraries that Python bindings to libsodium or NaCl (
​https://github.com/stef/pysodium or ​https://github.com/saltstack/libnacl or ​https://github.com/pyca/pynacl)?

[daira]

Replying to davidsarah:

The right way to construct a PK encryption scheme from the curve25519 key exchange is probably to use ​DHIES/ECIES.

Or better, crypto_box_seal.

[daira]

Replying to daira:

Why have this in pycryptopp rather than using one of the existing libraries of Python bindings to libsodium or NaCl (
​https://github.com/stef/pysodium or ​https://github.com/saltstack/libnacl or ​https://github.com/pyca/pynacl)?

This comment also applies to #91, #94, and to some extent #46.

(To be fair, I'm not sure the Python bindings existed when these tickets were opened. )