Having 2 sockets for loopback or multiple interfaces is a huge pain
Re:Why IPv6 is the only way forward
by BenjiWiebe on Wednesday April 08, 2026 @12:10AM
(#47685115)
Bind to '::' and you can get v4 and v6 traffic on one socket.
Doesn’t work for loopback (you have to listen externally )
It works fine for me on NixOS (Linux), with a recent kernel version and no weird config options - or at least I think so.
Not on all OSes.
And it's not all good to have one less socket, there are still issues with people not expecting v4 traffic on a v6-capable socket, like described here:
https://radar.offseq.com/threat/ipv4-mapped-ipv6-addresses-t...
If you want to handle two protocols, it is not unreasonable to use two sockets.
Why are you casting to sockaddr_in/6? The whole point of that system is that you can just pass around the sockaddr* without even needing to have a definition for sockaddr_in or sockaddr_in6. All of the socket API functions accept sockaddr*, and if you need to get the IP or port out then you use getnameinfo(), which also takes sockaddr*. There should be little reason to ever cast to either of those types in normal use. (I can think of one or two cases where you might, but they're not common.)
Having to deal with the separate socklen_t is mildly annoying, but you can just make a little struct that holds both.
10 years ago I was all gung-ho about IPv6, but it's annoying at every level.