On Thu, 16 Jan 2014, Branimir Rajtar wrote:
One more thing I find puzzling - RFCs 2473 and 6333 state that you have to encapsulate the packet first into IPv6 and if it exceeds the IPv4-in-IPv6 tunnel MTU, fragment the IPv6 packet. This draft states the opposite thing, if the encapsulated packet is larger than the tunnel MTU, you have to fragment IPv4 and then encapsulate it into IPv6. In my opinion, the simpler thing to do would be to encapsulate first and fragment afterwards.
You really really want to avoid having fragmented IPv6 packets on the wire if you can. Yes, I guess this might be ok since this is in within the provider network so they can control forwarding performance of fragmented IPv6 packets, but having the lwAFTR do re-assembly for most traffic is just bad for performance. We really really want to give guidance that avoids fragmentation regardless of where it happens.
I think this is also a problem because RFC 6333 describes fragmentation only from the AFTR's viewpoint and this draft describes it from the B4's viewpoint. Since this draft refers to RFC6333 for fragmentation, it would mean that the lwAFTR and lwB4 would have different fragmentation logic.
It would be good if there was clarification on both aspects I believe. -- Mikael Abrahamsson email: [email protected] _______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
