> > when there's no link layer address (imagine tunnel interfaces),
> > there's no proper NS (packets go out without NS-NA
> > exchange), however,
> > NUD hapens. is my understanding right?
>
> The MS implementation works that way. For a p2p interface, we create the
> neighbor cache entry in the stale state (since we know the link-layer
> address a priori), but then NUD can operate.
>
> Here's another scenario along these lines: assign a /64 to a p2p link
> between two routers. Now someone sends a packet to an address in the
> /64, but the address is not assigned to either router. The routers will
> forward the packet back & forth until the hop limit hits zero. This will
> happen before NUD has a chance to kick in.
>
> I agree with itojun, better to generate a
> destination-unreachable/address-unreachable error instead of forwarding
> a packet back out the p2p interface from which it arrived.
I'm inclined to disagree.
when pinging a local interface address for a p2p link, some IPv4
implementations send the ping out the link, and gets it back from the
remote router, as a way of verifying connectivity.
implementing your suggestion prevents the option of doing something
similar for IPv6.
let the packets bounce between the two routers for a while, in the end
you'll get a time exceeded message anyway.
/ot
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page: http://playground.sun.com/ipng
FTP archive: ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------
