I work for a device management company. I have devices running all over the world, potentially different ISPs. The problem is that we can't ping the devices using IPv4 because they are behind NAT firewalls (mostly). We can't have long lived TCP connections because the servers can't handle that many simultaneous TCP connections (could be 100000 clients at one time). I want to be able to send any of them a UDP packet (or TCP connection) to notify them to to start a session with the server (which has a known address/name). The clients would send a packet to the server to register their IP address with it. This would be easy in an IPv6 native world (or indeed a non NATed v4 world). I have been scanning these posts to see if I can use a solution from SisXS. I think not, but would be pleased with any suggested solution. As far as I see it I have several options. Firstly I could apply for a SixXS tunnel for each device. I think this would be too cumbersome. These are "dumb users" (don't tell them I said that, they don't know!), so the requirement is that we can't get them to fill in any useful information, let alone get them to give out their addresses. We can do any amount of code based configuration, just not interaction. So they can't have their own SixXS accounts. Secondly I could have a subnet, but that entails setting up a tunnel server endpoint (since the clients won't necessarily have an IPv6 network.) If I did this, I wouldn't actually need a tunnel broker, because the server doesn't actually have to talk to the rest of the Internet, just the clients. The problem with this approach is scalability. There could be hundreds of thousands, so one server wouldn't cut it and if we build a server farm then we don't actually need IPv6, we could just have the clients use a direct TCP connection to the server. Thirdly, more promising, Teredo. The target platform for the client for the moment is Windows 7. Using a SixXS account on the server (Solaris/Linux) would work, and in theory the Windows7 clients could connect using teredo and tell the server their IPv6 addresses. Problems with this as far as I can see: 1) The IPv6 addresses can change if several clients are behind the same NAT firewall, so the server would not know their addresses at any particular time. 2) I might well have read this wrong, but it looks like I would need my own teredo relay and or server for the sake of routing. This might be perceived as being a bit too complex to maintain. Please correct me if I am wrong here. 3) Most of what I read about teredo is that it doesn't work very well and there are routing issues. What I need is a way to auto-configure IPv6 hosts in an IPv4 world (like SisXS without any authentication. I can see why that is a non starter). Any ideas or corrections would be welcome.