Member
This. This right here. 10mbit is the sweet spot for "I need to check my v6-only mail server" and "no I will not torrent through this poor man's gateway." Feature, not bug, confirmed.
I'm on Telefonica in Barcelona, still no native IPv6 at home in 2024. NAT64 from pavel_train's box gets me to GitHub's v6-only raw content domains when my ISP's CGNAT fails. 10mbit is fine for git clone. Gracias for running it.
Following this thread because I'm building something similar in Melbourne on Vultr. Different problem: DNS64. Using bind with dns64 module, but AAAA synthesis latency is annoying. Anyone tried unbound vs bind for this? Not hijacking, just curious before I open my own can of worms.
Member
I use unbound in Curitiba for my home network, much easier config for dns64. The AAAA synthesis is fast enough, the problem is always the upstream v4 latency after, ne? The DNS is never the bottleneck, is the NAT that kills.
Pavel_train, you run your own DNS64 or you point users to Google DNS64?
Member OP
I run nothing | users bring their own DNS64 | I am gateway only | tayga layer | not service layer | separation of concerns | unix philosophy
Some use Google 2001:4860:4860::6464 | some use their own bind | some use unbound like nerdbann plans | I am pipe | not plumber
This reduces my attack surface | and my support burden | both are already too high
Been watching this thread. 40% drop rate is harsh but I get it. I'm in Dublin, most of the infrastructure I work with is dual-stack now, but I keep a Hostinger VPS in London for legacy v4-only monitoring endpoints. The idea of running public NAT64 gives me hives. Status page would go red, I'd get paged, I'd cry.
Hats off to pavel_train for the stubbornness. Not delusion. Stubbornness with metrics.
