I'm having to disable IPv6 to home for the time being, until I am Less Grumpy. I can use:
- HE.net tunnel: intermittent hangs and connection drops, makes long-lived SSH painful
- sixxs.net tunnel: MTU problems to a box at Cambridge University, used for Exim development, despite my having a 1280 MTU on my tunnel. Effect: connection hangs shortly after init.
- 6to4: sigh
6to4 used to work just fine for me, mostly. When there are glitches it can be hard to debug though, and it was a little irritating being on the West Coast of the USA and having all my traffic exit via a place in Germany. Fortunately, my colo-box is in The Netherlands, so that worked out fine, and actually had a slightly lower latency than native IPv4 connectivity. Route optimisation.
Alas, the lack of determinism in the reverse path means that, when I turn up 6to4 (East Coast, Comcast business class) now, I have lost return connectivity from my colo-box. The machine at Cambridge is fine. When I ping my colo-box, the ICMPv6 packets make it there, return packets are generated, but they don't reach me. The reverse path stops in telia.net. Administrative prohibition!
Comcast turned up 6to4 relays. That's nice. But peering spats are an area I just don't want to go into. *sigh*
UPDATE: I forgot that the diagnostic codes in traceroute6 differ from those in traceroute. The !A does not mean Administrative prohibition, it means address unreachable. Admin prohibition moved to !P — oops. Consistency, who needs it?