Re: [76attendees] Rogue IPv6 RA

[Erik Kline <ek at google.com>]

2009/11/9 Arifumi Matsumoto <arifumi at nttv6.net>:
> Erik,
>
> On 2009/11/10, at 10:43, Erik Kline wrote:
>
>>> If the latter paragraph only should be executed, the address given
>>> by rogue RA remains, right ?
>>
>> My reading would be that on receipt of a 0-lifetime RA that only the
>> second paragraph would be executed (lifetime timeout).
>
> Second to that.
>
>>  However, all
>> hosts receiving the 0-lifetime RA would then have to recompute the
>> next-hop, which in /some/ cases may require sending a RS (which the
>> rogue RA node would presumably hear and re-answer).  (Of course I
>> haven't verified this against any implementation :)
>
> I fail to get your point.
> Requiring sending a RS leads to ... ?
>
> Even if that RS fails, it does not have any effect on the given
> addressby rogue RA, right ?
>
>

I was contemplating the case where you might want to run a
rogue-ra-killer on a node that listens for RAs, knows which one are
valid, and sends 0-lifetimes for all the rogues.  My point was that if
a node decides it needs to re-RS then the rogue RA node probably
continues to reply (as does the rogue-ra-killer).  Maybe this doesn't
actually happen though because most/all nodes would have also received
the valid RA and would just update next-hop information using that
data.

Just me speculating idly on a weird situation...
-Erik