Re: draft-gont-6man-slaac-renum-01

Erik,

> On Mar 24, 2019, at 9:47 PM, Erik Kline <ek@loon.co> wrote:
> 
> To restate thoughts from IEPG this morning and from hallway
> conversations thereafter, I think it would suffice to add some
> "loopback detection".  Basically, if a host has cause to think a
> router no longer routes a PIO onto a link it should start some prefix
> unusability detection.
> 
> This would basically be a host sending a packet from itself to itself
> via the router.  The destination address should be a GUA from the PIO
> formed on the interface, but the contents and source address I'm less
> sure of (could be from link-local or the same GUA or a different
> on-link GUA, could be a ping or a unicast NS or a UDP packet to a port
> the host opens, …).

I think the challenge here is to avoid the hosts loosing a prefix that the host can use for local communication while there is a temporary problem with the router or the link to the router.  I can see a number of cases where the problem is the link to the router that at some point will be resolved.   As opposed that the router lost or forgot the prefix it had allocated.

> 
> If a positive packet reflection occurs then that router still routes
> the PIO onto the link. If there's no such response within $TRIES
> across $SECONDS, then assume that router doesn't forwarding that PIO
> onto the link.  (Note that the PIO might still be valid on-link as
> specified by a 2nd router…)

If the host has no other global prefix, I don’t see any reason to have the host stop using the prefix.

Bob

> 
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> ipv6@ietf.org
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------