Erik Nordmark wrote:
If you are probing to determine the initial PMTU, the answer is to send
periodic additional probes, of course. Routes are going to flap, and paths
are going to change - a once-probed MTU cannot be expected to remain
stable for any set length of time.
Did you really mean that? Clearly there has to be some assumption about the frequency at which the MTU can decrease. If not, for every e.g. TCP packet you send you would need to send an MTU probe (with an ACK) to make sure that a (TCP) packet of that size can make it to the neighbor.
No, we don't want to send an additional probe packet along with *every* data packet. If we can somehow use the data packets themselves as probe packets, then that would be a different story of course.
One case in which data packets could be used as probe packets is when IPv4 is used as an L2 media for IPv6. In this case, we could send IPv6-in-IPv4 encapsulated packets with the DF bit NOT set in the IPv4 header expecting that the decapsulator would send us some sort of indication if it sensed fragmentation.
But this begs the question of a fundamental design point: do we need to support sub-L2 media elements (i.e., the physical elements that sit below IPv4) that neither support IPv4 fragmentation nor send IPv4 "frag needed" messages when they can't forward a packet? Based on the 1Gbps/100Mbps Ethernet bridge example, I believe the answer to this is "yes" - would you agree?
Fred [EMAIL PROTECTED]
-------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
