bt90
bt90
@syncthing/maintainers any objections? I may have some time to invest in this in the near future and wanted to make sure we have a consensus before I start.
The comparison is a bit like apples to oranges. We do publish the public IPv4 address, albeit indirectly as @rdiez mentioned. While it's true that we don't announce public IPv4...
To us, this is done transparently, with the operating system managing the IP allocation and lifetime. It's not that different from IPv4, as your DHCP-assigned IPv4 also tends to have...
> Without being able to properly configure firewall rules, we're down to the odds of holepunching. to explain this point a little further: Of course, holepunching only works if the...
> About 25% of devices have an ipv6 address in the discovery server. The odds of any given pair of devices both having ipv6 should be about 6%, just statistically....
> However realistically compared to ipv4 where you have that tracking per subnet, that doesn't seem so much worse. I feel like users caring about that are users that shouldn't...
I think most NAT66 or NPTv6 deployments use addresses from the ULA range(`fc00::/7`). We already publish these as LAN IPs, but they are inherently not globally routable.
Isn't this something that quic-go should handle?
Maybe https://github.com/quic-go/quic-go/issues/1737 ?
Can you reproduce it or is this a once-in-a-blue-moon scenario?