4 ms·

Pure clickbait? The design choices are indeed bizarre, even if the reason ends up being for legacy support, no?

• corty 6y ago
  Well, bizarre would be using SHA3, some GOST algo, or maybe putting some postquantum crypto in there. The article describes something that is just a little strange.
  • minitech 6y ago
    PQC would be more interesting than bizarre and SHA-3 wouldn’t be bizarre at all (it’s a straight improvement on SHA-2 in isolation).
    • bawolff 6y ago
      > (it’s a straight improvement on SHA-2 in isolation). That's debatable.
      • tialaramex 6y ago
        Right, the only technical sense in which SHA-3 is simply "a straight improvement on" SHA-2 is resistance to length extension, but you can have that with SHA-512/256 (yes that's the name of one SHA-2 family member, if you didn't already know that, and you were previously confident SHA-3 is what you needed, time to do some reading). In a non-technical sense to be sure 3 is a bigger number than 2 but that's a terrible reason to do anything. Keccak's sponge construction is novel, it has some cool features compared to Merkle–Damgård but when SHA-3 was standardised we had very little experience with this construction. Now, maybe we lucked out and these are the perfect parameters, not too slow, not too weak. But most likely we went too far in one direction and so hence "Maybe Skip SHA-3" https://www.imperialviolet.org/2017/05/31/skipsha3.html https://www.imperialviolet.org/2017/05/31/skipsha3.html
• paxys 6y ago
  Writing a "perfect" encryption algorithm which doesn't actually support any of your existing clients or services would be bizarre. What they have done is standard software engineering.