👋 I realize this is a topic that requires a lot of discussion, and may not be fit for a small issue like this. However, I didn't see any other issue on this and figured it couldn't hurt to ask.
Currently this library uses Curve25519 which while fast and secure today is not secure in a world where quantum computers exist (note: generally NIST/most organizations seem to be planning to deprecate non post-quantum safe algorithms in 2030, which is still a couple years away. However, it can take time to roll these out so it might be beneficial to talk about them now, or if anyone is concerned about "store now decrypt later" attacks). So I was hoping to kick start this topic with this issue.
Would there be any interest in reviewing contributions that provide some APIs to use something other than Curve25519 (e.g. a build time option to opt into using PQC algorithms, and not worry about migration right away. To start testing how it impacts the protocol)? Is there a larger plan to tackle it later? Is it something yall aren't considering right now?
I have many more questions about the use of Hybrids vs ML-KEM, and many other things, but don't want to blow up the discussion too much right away 😉
👋 I realize this is a topic that requires a lot of discussion, and may not be fit for a small issue like this. However, I didn't see any other issue on this and figured it couldn't hurt to ask.
Currently this library uses Curve25519 which while fast and secure today is not secure in a world where quantum computers exist (note: generally NIST/most organizations seem to be planning to deprecate non post-quantum safe algorithms in 2030, which is still a couple years away. However, it can take time to roll these out so it might be beneficial to talk about them now, or if anyone is concerned about "store now decrypt later" attacks). So I was hoping to kick start this topic with this issue.
Would there be any interest in reviewing contributions that provide some APIs to use something other than Curve25519 (e.g. a build time option to opt into using PQC algorithms, and not worry about migration right away. To start testing how it impacts the protocol)? Is there a larger plan to tackle it later? Is it something yall aren't considering right now?
I have many more questions about the use of Hybrids vs ML-KEM, and many other things, but don't want to blow up the discussion too much right away 😉