From my searching it seems a tradition to ask each new year how IPv6 UDP support is coming along!
I am exploring ways to host a DNS service and having IPv6 is (these days) pretty wanted for that kind of thing.
I can see and read that basically the only use case for this is DNS (mostly) and it’s probably not high on the list for Fly, but after a few years of on-and-off news I hoped it would be along by now
Hope we can get another update on how the priorities are on this!
QUIC won’t be easy. You’d need a QUIC-aware load balancer / reverse proxy as the protocol supports connection migrations across servers and client IPs.
I’m hoping to establish stateful WebRTC connections with individual Machines, and I’ve hit a bit of a roadblock. Without UDP over IPv6, I’d have to allocate a dedicated IPv4 address per Machine (i.e. possibly spin up a new app per each Machine), which rapidly becomes expensive at $2/mo for each IP.
Or perhaps there’s a better way to accomplish this?
I’m also looking at Fly.io’s suitability for hosting a DNS service. In my case IPv4 is more of an impediment than a benefit and ideally I’d have a unique IPv6 address per user. Hopefully this is something Fly.io can fulfil at some point in the near future.
wait Public Network Services · Fly Docs seems to indicate we can control anycast for UDP, sou you could run your own DNS anycast and other things like http3 / quic ?
+1 for IPv6 UDP! I have a case where I want to route UDP traffic (from a client I don’t control) to specific single-tenant machines. The client supports IPv6, so this would be a very easy way to direct them to the right machine, and without that I have to either put each machine in its own app with an IPv4(!) or run each machine on a different port, tracking which machines are consuming which ports and providing that to the clients.
With Fly.io starting to charge dedicated IPv4 addresses, and shared IPv4 addresses not supporting UDP traffic, UDP support on IPv6 will surely be of great help to developers.