Having read up a bit about ethernet jumbo frames, I'm wondering. There seem to be technical advantages to being able to use larger than 1500 byte ethernet frames in some cases. Just enabling jumbo frames on a network card is not backward compatible and should not be done unless the entire network supports jumbo frames. I've read the IETF draft and I'm wondering why there is no autonegotiation for jumbo frames.
The latency of 9K frames on gigabit ethernet is less than that of 1500b frames on 100mbit, and technical problems such as the CRC being inadequate for large frames could also be fixed by autonegotiation. Mixing jumbo and non jumbo ethernet equipment would require a solution such as ethernet-level 'fragmentation required' messages or separate mtu discovery for every destination mac address, but it doesn't seem unsolvable.
- Other backward incompatible ethernet settings are autonegotiated, why not jumbo frames?
- Are there technical reasons jumbo frames could not be autonegotiated?
- Are there political or other reasons?