In reply to a post by omegapoint6:

I’m no longer receiving an IPv6 address, no DHCPv6 response from what I can tell

All ok here still...

In reply to a post by robsdesk:

In reply to a post by omegapoint6:

I’m no longer receiving an IPv6 address, no DHCPv6 response from what I can tell

All ok here still...

Figured out it was a firewall rule somehow blocking the DHCP renewal

Also heard from cuckoo support, they assign a /56 block & for static IPv4 customers the prefix allocated is also static.

In reply to a post by omegapoint6:

Also heard from cuckoo support, they assign a /56 block & for static IPv4 customers the prefix allocated is also static.

The IPv6 prefix should be static for all customers. Both Windows and Linux clients take 2 IPv6 global addresses. One static and the other dynamic, so the end user can choose whether to go static or dynamic on a client by client basis

Another ex-Swish refugee now part of the happy Cuckoo flock here. All my kit has been IPv6 enabled for yonks, and a few weeks ago I noticed that the router and my motley collection of Android, Windows, & Linux boxes had acquired themselves IPv6 addresses starting 2A0B: and were merrily using them (the router's is a /64). I could ping the router's IPv6 address, so I set up a BQM to it, something I haven't been able to do since I went to FTTP with Swish and got stuck behind CGNAT. It all worked fine; everything looked rosy.

Then this week something broke. It was about the same time that we had a short power cut here, which muddied the water a bit, but I think that was just coincidence. BBC apps on Android - News, Weather, iPlayer, Sounds - all became extremely slow to the point of being unusable. The same pages worked fine on Windows/Firefox, and there were no reports of issues, DoS attacks etc that I could find. test-ipv6.com gave the connection 0/10. The router still had its IPv6 address (via DHCPv6?) and the BQM still worked.

Long story short, I have had to disable IPv6 support on my router, and now the BBC apps on Android work fine again. I'm guessing that Cuckoo must be testing their IPv6 stack, and they have broken something that Android/BBC doesn't detect, so just sits there waiting for IPv6 to timeout. Other sites seem to fall back to IPv4 more quickly. Windows works because bbc.co.uk is NODATA for DNS AAAA requests, so falls back quickly. Does bbci.co.uk (used by the mobile apps) have an IPv6 address in DNS? It's not so easy for me to check this now.

I run my own TP-Link router. Possibly that ISP supplied ones have had IPv6 disabled to avoid problems like this?

As an aside, I have made several calls to Cuckoo support, and to say that they are diabolically awful would be extremely flattering. Even Thames Water are better. It is all outsourced to people for whom is English is not their first language. None of them is techy, and they will not transfer me to someone who is. They will only write down my questions as best I can make them understand, then call back with their understanding of the "answers". You can imagine the quality of the answers resulting from this process.

After a few issues with my account being migrated from the old Swish platform to the new APFN Aquila platform, my IPv6 is working correctly (note that you need to be migrated for it to work). A few data points for others:

1. You'll get a /56 via DHCPv6 prefix delegation (range is in the APFN 2a0a:5583::/32 block, AS5482)
2. I don't get an address via IA_NA but that's no issue, the link-local fe80::/10 can be used as default route.
3. As with IPv4, the lease times are short, 5 minutes.
4. For those using OpenBSD, everything works fine with dhcp6leased(8)/rad(8).
5. Cuckoo is my retail ISP and they are not great. The staff try their best and are helpful but there's no way to speak to anyone other than L1 support.

Note that I have a static IPv4 address, it may be different if you're behind CGNAT.

Edited by mjhturner (Mon 16-Mar-26 11:07:42)

Following @mjhturner's post I re-enabled IPv6 on my router, and set up the v6 BQM again. It's been running for 2 weeks now, and all seems to be well. The bbc apps on Android are fine, and I get 10/10 from test-ipv6.com. But there's been no official announcement of the service from Cuckoo/APFN, which makes me a beta tester I suppose, and I've been here before, so no chicken census just yet.

Re @mjhturner's comments and caveats: I don't have a static IPv4 address and I am behind CGNAT (but it's not clear to me why this should make any difference in v6). Not sure I've understood the IA_NA point, but my TP-Link router has both a WAN side address and a LAN side address, the latter SLAAC generated, both starting 2A0B: but from different address blocks. My devices all have SLAAC GUAs using the same prefix as the router's LAN address. I deduce that both the IA_NA and IA_PD must be working in my setup. I can see RA messages including the prefix with Wireshark. My modest router doesn't expose details of what size the delegated prefix is, /56 or whatever, nor what its lease times are. Everything on the LAN downstream from the router is /64, as expected. The router logs are normal - it doesn't seem to think anything extraordinary has happened.

Address prefix lease times (preferred and valid) are 5 mins on my Windows and Linux boxes, and also in the RA frames. Likely it's the same everywhere (Android is not so easy to check) and the router is simply passing on values received with the prefix from APFN. Typical lease times quoted on the Web are hours or days; IPv6 should have obviated the need for short leases provoked by scarcity of v4 addresses. Maybe it's a sign of APFN's lack of confidence in their stack, such that they might need to reconfigure at short notice? Despite this, addresses seem stable over time, router reboots etc.

My router pushes out an ICMPv6 RA on average about every 7 sec, which seems excessive as the payload doesn't change. I've seen 200 sec suggested as a more usual interval. I've never run IPv6 before, and have nothing with which to compare it. Is it maybe a consequence of the short prefix lease times, because address refresh needs to happen often? With a longer interval, new devices joining could always send an RS to FF02::2 to get going?

Address blocks starting 2a0b: are part of AS5482 as well as 2a0a: (and others). A quick back-of-envelope sum suggests APFN have ~(2^99 + 2^96) or ~7.1 * 10^29 registered addresses. This is only about 1/10^9 of the total v6 address space. Boggle.