On Tue, 21 Sep 2010, Manfredi, Albert E wrote:
I think your idea makes most sense in cases where Internet access is provided via cable head-ends or PONs. Where everything from customer premises has to go through the head-end anyway. Yes, the need for ND in this particular case is probably questionable, at least within the provider's coax network or PON.
In L2 ETTH networks it's done the same way, by means of "local-proxy-arp" and disallowing hosts to send arps to each other.
Otherwise, I don't see the advantage of overloading the DHCP server with this new responsbility of keeping track of all physical addresses in the local net, other routers, etc. As Hesham pointed out, this concentrates more responsibility on a central server, a responsibility that is now distributed among all of the hosts (and routers) instead.
I'm perfectly fine with the router itself being DHCPv6 server.
My position is, I need to have a reliable way of distributing "well known" IPv6 addresses to the hosts of peer-peer networks. That's my interest in DHCPv6, vs SLAAC, for assignment of addresses.
Well, it's not my interest. I want tracability and security, and I can only do that if my customers only communicate to each other via L3 via my router. I never want customer to talk direct L2 to each other. Q-in-q is expensive to terminate so I still want them to be in the same vlan.
-- Mikael Abrahamsson email: swm...@swm.pp.se -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
