Hi Woj,
On 10-09-08 04:16 PM, Wojciech Dec wrote:
Sending periodic RAs with the PIO does not help with the two
problems that were pointed out:
- the network does not necessarily know when a host attaches,
because the host may timeout sending RSs before the link layer
is available to carry these RS's up to the node assigning a
prefix. As a result the network does not know that a host
attached, does not allocate a prefix, and cannot send a periodic
RA that contains the PIO necessary for that hosts to complete
SLAAC. The hosts ends up with no connectivity.
Correct. There needs to be some work done in BBF to convey the
information about the line coming up from the AN to the Edge Router.
Well, actually that work (line up/down info between AN and edge router)
has already been done (and in the IETF too) by the ANCP WG. That however
by itself is not sufficient, in that there needs to be a destination
address of a data plane device eg the MAC of a CPE, to send the RA
packets to. An alternative "solution" would naturally be to throw away
the regular ethernet forwarding paradigm on an between the Access-Node
and Edge Router by inventing and deploy a new dataplane based around
forwarding of line-id addresses. At this point one could well say that
the cure is worse than the problem it tries to solve...
Thanks for the pointer to ANCP. It was very useful. I also don't think
that it makes sense to invent a line-id based dataplane between the Edge
Router and the AN. But I think it is entirely viable to do line-id based
filtering on the AN.
Cheers
Suresh
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------