Re: Confirm consensus to adopt draft-hinden-6man-mtu-option

[Ole Troan <[e-mail]>]

David,

> I agree that the general state of PMTUD on the Internet is poor, but there are limited domains where the state of PMTUD as currently specified is mostly functional, the R&E world is one of those domains, and with the scope defined for this draft it is quite possible to implement such a domain with functional PMTUD as currently specified. So, part of the experiment needs to compare this new solution to the current solution in limited domains where the current solution is functional, especially if this solution is to ever go beyond the experimental stage.
> 
> Furthermore, the end-to-end use of larger MTUs is an end-to-end problem and focusing on PMTUD to the exclusion of the first and last hop LAN problems is doomed to failure. The point is even if all routers between two hosts implement this HBH EH perfectly if large MTUs (>1500) on the LANs at each end are not specified and interoperable how is this going to work? I'm not aware of BCPs for Large MTU LANs and I don't know of IPv6 test suites that test the interoperability of hosts on Large MTU LANs.
> 
> So if you want to do 9k MTUs between Data Centers we need a definition for 9K MTU on LANs between hosts that is consistently interoperable, which I don't think exists.

As a side-note it might be worth looking at https://tools.ietf.org/html/draft-van-beijnum-multi-mtu-05
One could imagine using the HBH MTU option in NS/NA exchanges between directly connected nodes too.
But that would still only be a hint, probing would be necessary.

Cheers,
Ole