On Tue, 8 Sep 2026 19:13:40 +0000 Lee <[email protected]> wrote: > > > Aren't applications _supposed_ to try ipv6 first and, if that doesn't > work, fall back to ipv4? >
If both are apparently available, yes. But if one is disabled in the kernel, it doesn't make much sense to try it at all, does it? And it doesn't always seem to work. I had no address found for a repository which I knew worked, using bind9 with root hints. Dig found it fairly easily using the router's resolver, which I try to avoid doing, but couldn't find it at all using bind. It took me a while to realise that bind was only trying IPv6, and definitely not falling back to IPv4. But it was working properly for everything else, just that one repository, which definitely did have an IPv4 address. A suitably IPv6-disabled bind now finds it easily. An application can't know if the router(s) have IPv6 disabled, or the ISP's router outside, but it can find out fairly easily whether its own kernel has it enabled, either on the fly or a flag can be set when the application starts. -- Joe
