Erik, > -----Original Message----- > From: Erik Nordmark [mailto:[email protected]] > Sent: Tuesday, December 08, 2009 9:06 AM > To: Templin, Fred L > Cc: Hemant Singh (shemant); [email protected]; Pekka Savola > Subject: Re: speaking of ND Proxy and NBMA etc. > > Templin, Fred L wrote: > > > Some links rely on unicast RS/RA and not multicast. Wouldn't > > unicast-only proxy ND avoid the looping issues you were > > concerned with? > > No. > > If the network is misconfigured you can get loops.
Then, let's be sure the network is not misconfigured. Can't we use spanning tree or some such to ensure that? Fred [email protected] > An example: we have > host A sending (even a unicast NS) to the L2 address for router R1 for > target B, that causes R1 to send (even a unicast NS) to what it thinks > is the L2 address for B, but that turns out to be an incorrect L2 > address. For instance if that L2 address is that of some other router R2 > that does proxy ND, then it might in turn originate a NS for B send to > the L2 address of R1. Then we'd have R1 and R2 sending NS packets for > target B to each other at full line rate without any limit. > > If some of the NS are multicast the risk gets worse, because the looping > can be combined with the multicast duplication. In the example, > multicast might mean that R3 and R4 would also receive an NS for B and > decide they need to send a NS for B to R1. > But the unicast looping is bad enough. > > Erik -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
