On Sat, Aug 8, 2020 at 5:41 PM torix@protonmail.com wrote:

An important piece of mental furniture that I never needed to know until months later when a relay stopped working is right here in Step 7: Check IPv6 Availability of page https://community.torproject.org/relay/setup/post-install/ "If you enable IPv6 without working IPv6 connectivity your entire relay will not be used, regardless if IPv4 is working." This seems to me to be quite idiosyncratic to tor, and easily forgotten - not the first thing I would think about when troubleshooting, nor is the post-install page quite where I would look for help when a relay stopped. As we move more and more to including IPv6 in our relay setups, this will come up more often. But we should at least add bold here. --Torix

Yeah, I would happily take patches as needed, to the code or the documentation, to make this issue more straightforward. (I don't maintain the website, but I bet they'd take patches too.)

The design issue here comes down to whether you'd rather have your relay fail silently or fail loudly. My understanding that the last time we asked relay operators about the issue, the more popular opinion was that they'd rather not have their relays silently ignore non-working parts of their configuration.

Medium-term, with the 0.4.5.x series, Tor will do a better job of detecting whether a configured IPv6 address works, and a better job of detecting whether you have an IPv6 address if you don't specify one. If you explicitly set an IPv6 address, it has to work or your relay won't publish. But if Tor infers an IPv6 address for your relay, your relay will ignore the address if the inferred address doesn't work.

This might all change in the future: generally, it's a lot safer to take a formerly non-working configuration and make it permitted than it is to take a working configuration and disallow it.

cheers,