> Throw in IPv6 traffic and suddenly each compare takes twice as much time in the CPU So an operation responsible for less than 0.1% of the total time spent ro

> Throw in IPv6 traffic and suddenly each compare takes twice as much time in the CPU So an operation responsible for less than 0.1% of the total time spent routing a packet now takes twice as much time... or it would if CPUs weren't superscalar for a long time now. That's paid back in tenfold just by the omission of the header checksum. Do you have any benchmarks showing a real-world difference? https://www.unix-experience.fr/2013/ipv4ipv6-performances-comparison/ https://www.unix-experience.fr/2013/ipv4ipv6-performances-co... shows performance to be identical on modern computers.