Since those experiments, linode is now also providing native IPv6 so my linode was switched to the auto-provided address they provide. By default they only provide 1 IPv6 address but they allow rDNS to work, so I haven’t needed any more, yet!
IPv6 speeds internationally are a lot faster as well. I did some speed tests to a site in a UK exchange (connected at 100Mbit/s). It would saturate my home FIOS connection, peaking for periods of time at 3.65MB/s. When I attempted the same speedtest from my linode it was quite variable and bounced between 1 and 4MB/s, and was averaging slower than my FIOS. But when I tried it via IPv6 from the linode the average was 7.42MB/s with peaks up to 9.8MB/s. So here we see the IPv6 connection running over 3 times faster than the IPv4 connection! Funky stuff.
My home tunnel is now endpointed on a Seagate Dockstar, which I’ve got running Debian. The Dockstar is a “pogo-plug” type device, and has a Gbit network port. I’m not sure how fast the machine can drive it, but it’s definitely fast enough for IPv6 traffic to the outside world. This machine runs the tunnel and ip6tables. Now what this also means is that my linode and v-colo can talk “directly” to my home machines. So I no longer need my UUCP over SSL method of sending email; my home machine can talk SMTP over IPv6 to my virtual hosts (redundancy!) and those hosts and send mail directly to me. Faster, more efficient, more resillient.
What I do need to work out, however, is “fixing” the tunnel if my IPv4 address changes. HE tunnelbroker provide a web page to allow automatic updates, but they don’t like you hitting it for no good reason. And I can’t find an automatic way of reading what the IPv4 endpoint address is. So once or twice I’ve had my IPv4 change and so the tunnel freezes. I need to work on fixing that!