On: Predicting and controlling NetHack's randomness (2009)

[vinceguidry]

I imagine that would make the server pretty slow in times of heavy load.

[wolf550e → vinceguidry]

You can use 24 bytes of "true" randomness (16 for key and 8 for nonce) and then get 2^64 16-byte blocks of randomness before re-keying at better than 1Gbps per core using AESNI. How much more do you need?

EDIT: How is what I wrote wrong?

[tptacek → wolf550e]

It's not wrong, and I didn't downvote you, but I'm not sure we need to concede that anything more difficult than "read numbers from /dev/urandom" is required here. :)

[tptacek → vinceguidry]

I doubt it. Things exercised far more strenuously than nethack rely on urandom.