Hacker News Reader

Even if the data is not random it doesn't weaken your kernel prng by writing non-random data to urandom.

Don't overcomplicate your threat model with non-existent risks

Didn't DJB somewhere describe a PRNG model in which this is not true and you shouldn't let an adversary give you PRNG seeds?

(I don't have any problem with what CloudFlare is doing.)

Edit: While I was writing this comment, drewbug linked to the exact DJB post I was thinking of. Thanks!

Someone observing the lava lamps still can't control their appearance.