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

[Philip Homburg <[e-mail]>]

> The only "right thing" we need routers on the Internet to do is
> too ignore HBH EH instead of dropping packets with them or relegating
> packets to some slow path. 

First, there is an obvious deployment problem. It is not attractive for
hosts to send HBH EHs if they get dropped on a large percentage of the 
paths. In the short run this will seriously complicate host processing.
In addition to ICMPs, there will be dropped EHs, ignored EHs, and remote 
hosts that don't know what to do.

At the same time, it is very costly to build fast path processing for HBH EHs
(assuming there is none at the moment) only to support some future internet
where everybody will support this option.

And then we can safely assume that some percentage of paths, so 5% or 10%
will continue to do the wrong thing. There is no way you can get this in
every last firewall.

That said, if we find a way to define this for PMTUs >> 1500, then it is
easier to say it will only work if all devices on the path support it.

Note that there is also
https://tools.ietf.org/id/draft-van-beijnum-multi-mtu-04.txt
to deal with the lack of support for >1500 by IEEE.