Re: [v6ops] New Version Notification for draft-palet-v6ops-p2p-from-customer-prefix-00.txt

[Ole Troan <[e-mail]>]

Mark,

I most definitely think doing address resolution on links without link-layer addresses is the wrong thing to do.
Sure, other ND functions are being used. NUD isn't typically useful on router to router links. Often p2p links have link-layer notification on link state too, and I'd argue that the benefit of a router doing NUD towards hosts have limited usefulness, unless it is to garbage collect state.
If we ever get to update NUD, I would argue quite hard for simplification of the protocol.

> There are two choices. Either prevent the packet loop occurring if a
> packet is sent to a destination that doesn't exist, by probing for its
> existence via ND NS, or mitigate the consequences of the loop after it
> has already occurred. I think prevention is better than mitigation.

If this is the problem you are trying to solve, then we have already solved that.
See rfc4443, from changelog:

    - In the description of Destination Unreachable messages, Code 3,
     added rule prohibiting forwarding of packets back onto point-to-
     point links from which they were received, if their destination
     addresses belong to the link itself ("anti-ping-ponging" rule).

Was there any other unresolved problem I missed?

Cheers,
Ole