IPv6 Whch range is allocated to the WAN port?

Quick question for any folks running dual stack;

Which address range is your router's WAN port being addresses in? The /64 ND or or allocated /48 PD

In reply to a post by IamQ:

Quick question for any folks running dual stack;

Which address range is your router's WAN port being addresses in? The /64 ND or or allocated /48 PD

Based on my current setup.

The router's WAN port (a PPP interface on xDSL) receives the /64 address using wan-network-prefix::1
According to the Zen settings this is using DHCPv6 over the native v4 connection.
(Fritz!Box)
The router configures its LAN network using the 0th subnet of the /48 PD (lan-network-prefix)

It assigns itself an address on the LAN based on lan-network-prefix::host-part
The host-part is based on its own LAN interface MAC (EUI-64), and the same host-part is used by the LL ( fe80:: ) address.

Client devices on the LAN typically see the default route as via the LL address of the router.

All fairly standard as far as I can tell.

Thanks. All the connections I've seen have used the bigger /48 to address the WAN port. Must be a config thing

There are several options, these are discussed in https://www.ripe.net/publications/docs/ripe-690 section 4.1

Well these are the settings automatically provided the Zen ISP profile selected on the Fritz!Box rather than me tweaking the WAN in any way.
I guess it might be different for people on FTTP or leased lines.

That said having read the RIPE document I think the non-overlapping /64 link for WAN and then separate LAN prefix delegation of /48 (or possibly /56 for home consumers) is simpler for compatibility and troubleshooting.

While many home users won't need more than a /56 I would prefer ISPs ask at sign up (with a default for those who have no preference or awareness) and not think of addressing schemes as having a role in market segmentation, and this is hinted at in the RIPE docs too, as it suggests being wary of applying a v4 address conservation mindset to v6.

I wonder what examples you have seen of the WAN link being assigned from the /48.
I've only had that experience with a leased line provider where an IPv6 allocation had to be explicitly requested, and even then it too them 3 different attempts to setup their upstream router to actually route replies back to me.
Their 3rd attempt settled on a /127 for the WAN link even though a /64 is usually more straightforward.

However the IDNet and Zen allocations just worked automatically including the prefix delegation suggesting is more a question of ISP experience at scale than the cost of the service.

Edited by prlzx (Mon 06-Feb-23 21:51:59)

Just had a look at a couple of Draytek's on FTTC. The WAN looks like its got a /128 on closer look, at the bottom of the /48

IP's masked with x's

WAN1 IPv6 Status:
2A02:80xx:xxxx::1/128
FE80::1/128

LAN Status:
2A02:80xx:xxxx::2/48 (Global)

I'll see if I get a chance to check any others over the next couple of days.

Are those Xs the same numbers, they might only be slightly different.

/128 typically denotes a PPP connection - like with the v4 /32s there's no subnet or ARP so the peer IP can be anywhere in the ISPs ranges.

On the LAN side I assume it is not really handing out a /48 prefix to end user clients but an nth subnet of standard /64 size.

Edited by prlzx (Tue 07-Feb-23 13:17:54)

Yep same numbers where the x's are, and the LAN side just routes the entire /48 for other internal things. Ironically not had any reports of issues.

The IPv6 is setup as PPP and when the PPP goes down internally the v6 still works (Unlike other's who have said if the PPP goes down they can't route anything inside there networks on v6)

Interesting results. The LAN-advertised /48 prefix will have a lifetime such that clients may persist over a brief outage and the router might not stop advertising it right away either, only to say that it temporarily doesn't provide a route for ::/0

btw I prefer where routers support options for advertising a ULA prefix alongside that can persist independently of a prefix received via WAN becoming invalid (or dynamic).
I think that's preferably for establishing a stable internal DNS too.

Fritz!Box has the ULA option and it should be trivial on something like pfSense, VyOS, EdgeOS and presumably OpenWrt.
(I do wish AVM would allow us to set the private internal domain though, which is why I have something else on the LAN doing the internal DNS. But that's off topic.)

For people with a single internal network, if the items fully support dual stack and are mdns aware, service discovery may continue to work over the link-local addresses anyway.

Edited by prlzx (Wed 08-Feb-23 02:13:12)

Just had a quick look, and on that box RA's are disabled and most things are addressed statically on the v6 side. There is a separate DHCP6 server handing out addresses to a handful of clients.

Everything has a persistent route to ::/0 via 2A02:80xx:xxxx::2 which is statically assigned to the LAN interface on the router so that's why they always stay working. Wonder what happens with the DHCP6 clients though as there is no default route sent with those requests I suspect they fail.

I'll see if I can do some testing at some point.

Edited by IamQ (Wed 08-Feb-23 21:24:44)