Design need of checksum fields at multiple "TCP/IP" model layers ? are they really redundant ?
let us take one example,
in L1: Ethernet-II, we have 4 bytes FCS field, in L2: Ipv4 we too have checksum, but in IPv6 they have omitted checksum and off-loaded to upper layers. in L3: TCP/UDP both mandates checksum field and further cause performance delay.
So, if IPv6's design to offload the checksum to uppper layers was a good design, then can we ignore L2 and above checksums as L1 FCS (frame check sequence) has already covered the integrity ?