4 ms·
A Tale of 2 Protocols: IPv4, IPv6, MTUs and Fragmentation
- pmjordan 18y agoWow, that was long. Interesting, but excessively verbose. I'll attempt a summary: Larger (~1500 bytes gross) IPv6 datagrams are currently not being routed correctly over certain networks. The author has deduced that a combination of multiple factors is to blame: - IPv6 fragmentation happens at the source, if a packet attempts to cross a link with an MTU that is too small, an ICMPv6 packet informs the sender. - IPv4 fragmentation typically happens on the way, as necessary. TCP supports an optional IPv6-like source fragmentation mode which isn't widely used as ICMPv4 packets are often filtered by firewalls, etc. - IPv6 coverage isn't perfect yet, so there are IPv6-over-IPv4 tunnels in place in many cases. Tunneling adds size to a packet. - Some tunnels refuse to do IPv4-based fragmentation for performance reasons. Because of the extra bytes, most packets would have to be fragmented. Instead, they use the TCP/IPv4 source fragmentation mode, but the relevant ICMP packets get lost along the way, so the tunnel can't operate correctly. Packets are eaten. I hope I've understood & summarised it correctly. If anyone with more patience or understanding of these things could correct me, that'd be great.