When setting a big MTU size in the LAN (say IP MTU 1500), it may produce fragmentation in the LAN (say by adding a GRE Tunnel). Currently, the WAN-model avoids fragmentation and calculate the right MTU Size for the UDP layer. As the WAN-model gives no feedback to the LAN-model, the WAN-model gets another MTU for the UDP layer, and both models are out of sync.
Options:
- reduce the MTU size in the LAN
- cope with fragmentation in the WAN, and calculate accordingly
- warn the user but do as if we have jumbo frames in the WAN