I have two external connections to two different ISPs. They're both connected to a router/firewall running OpenBSD. With IPv4, I have to do NAT on the router and I can do routing based on source address, destination port, load balancing etc. in the PF packet filter rules, which works at once for all my internal machines (mostly Linux machines). Is there a manageable way to make such a setup for IPv6 too? I now have a tunnel and subnet from one of my ISPs, and on the other connection a (static) SixXS tunnel - and about to ask for a subnet. As far as I understand, SixXS won't route packets originating from the ISP's subnet (and I guess the other way around, but I didn't yet ask). So if I'd have two subnets, and advertise them both to my internal network, the routing problem already starts in the internal computer: it somehow chooses one or the other address and thus determines which tunnel to use. Routing the packets in the router is no problem, but the router cannot make this decision: the originating computer already has to do this. Do you have any thoughts how to solve this problem? I could come up with the following ideas: 1) Could 2 SixXS tunnels - one over each outgoing connection - to the same PoP be a solution?Would the PoP be willing to accept packets from one subnet over two different tunnels? However, if both tunnels have the same IPv4 address at the PoP, it still won't work as I then can't route the IPv4 packets encapsulating the IPv6 packets. And I guess that this typically is the case :-(. 2) I read Philipp Kollmann's post which gives some extra possibilities of source-based routing on Linux, but not really things like load-balancing or destination port based routing - if I read the man-pages correctly. Moreover, it means setting up things in each internal (Linux) computer. 3) Is there any way that routing protocols like BGP can help in advertising routes to my internal computers? I'm really a newbie in this respect, so I'm just asking...