Re: [v6ops] draft-van-beijnum-multi-mtu-05

[Philip Homburg <[e-mail]>]

>Gert Doering wrote:
>> I totally see this as being great fun for IXP routers.
>> 
>> The draft really only seems to be talking about end-system talking to
>> each other, but for routers, it's quite a bit more complex.
>
>it's not just IXPs routers: the protocol adds failure modes at the
>bottom of the protocol stack, some of which will end up with packets
>being blackholed.  Operationally, it's a good idea to avoid situations
>like this:  apart from trashing connectivity, problems relating to
>network layer blackholes are troublesome to debug and fix.

Assuming two nodes (hosts or routers) have announced such a larger MTU. And
they have verified (using ICMP, or whatever) that is is actually possible
to exchange packets at a bigger MTU.

Then what problems do you expect in todays networks?

Do you expect bridges to corrupt packets even though the test packet work out.
(And is a there a class of devices known to do this)

Do you expect a spanning tree protocol to pick a route through different bridges.
(Does anyone expect to deploy ethernet jumbograms in such an environment)

Or maybe other source of failure.

I assume that to be safe, this option should be disabled by default and would
require explicit action by an admin to be turned on.