On 24-03-11 4:30 PM, Victor Lyapunov wrote:
Hello
I have been testing some scenarios for IPv6 over broadband
connections. The setup is a the most common one, the CPE gets
-One ::/128 WAN ipv6 address using autonegotiaton.
-A signle ::/56 LAN subnet for the user networks, through DHCP-PD
(further subneted into /64 subnets for the various VLANs in the CPE)
For this setup the NAS server is configured with a local ipv6 pool for
WAN address assignment (autonegotiation)
ipv6 local pool PPPOE 2001:100::/64 128 shared
And a second pool used by the DHCP_PD
ipv6 local pool LAN 2001:200::/48 56
In this way I have to maintain two different pools (one for CPEs WAN
and one LAN addressing).
A possible alternative that is discussed, is having the NAS allocate
just the DHCP_PD ::/56 prefix to the CPE (as far as global addresses
are concerned). And then configure the CPE to use the first of the
resulting 256 ::/64 subnets for the WAN and the rest for the LANs.
What is your experience, is the second alternative worth pursuing? Is
there a common practice?
PPP for IPv6 doesn't need a global scope address. It doesn't even
negotiate one, the only thing that is setup by IP6CP is link-local
addressing.
In our setup, we have no address for the WAN link, we just delegate a
/48 via DHCP-PD to the CPE.
The CPE can then decide to put the first /64 on the WAN link, if it
wants, but it doesn't have to (the CPE we use, AVM Fritzbox, does do
exactly that though).
The advantage is that you *know* what address or range the CPE has. If
it also uses an address gotten via SLAAC on the WAN interface, how are
you going to find the customer when the government asks "who is (or was)
behind this IPv6 address" and it's from your shared pool ?
Mike.
_______________________________________________
cisco-nsp mailing list [email protected]
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/
