If you’re trying to use a tunnel of some sort, or say for example IPv6 over IPv4, there is always the risk of problems due to reduced MTU. Long term, could there be a fix for this? I have a vague memory of reading something about BT having an MTU of 1600 for protocol x (unknown) - not necessarily IP, but I can’t remember the details nor where I read it.
It seems to me that there are two ways to fix this - one would be to propose a standard reduced IP MTU well below 1500 for applications’ routine usage; maybe 1400 but that might not be low enough so perhaps the already accepted figure of 1280. That way adding tunnelling headers would not break anything, just bring the total xPDU MTU nearer upwards towards 1500 which is bound to still be fine. A weak example of this is IP MTU 1492, as everything works with this I’m fairly sure. IPv6 MTU 1280 certainly works as a far lower universal MTU because it’s in the standards.
An example of sinful kit is my 3G ‘dongle’ NIC, which has an IP MTU of a mere 1440 iirc.
The other far more expensive option would be longer term to try to get internet routes to have an IP MTU of say 1600 or higher and that figure would have to be even higher for the L2 PDU MTU, say for ethernet frames. If you’re going to increase it and break things, then it might be well worth increasing the IP MTU by quite a bit more than 100 bytes for IP, maybe increase it a lot. This is far less realistic but it might happen if there were some policy document and a long ‘deadline’ or ‘go date’. If purchasers gave notice that they want jumbo frame, larger super jumbo support or a much larger still - who knows what - MTU supported on all new kit by a certain date maybe. I think BT is already doing the right thing. After all, I can send baby jumbo IP packets already with PPP MTU 1508 over BT21CN right now.
I thi