Would a PR be accepted that adds "dual-stack" TCP/UDP listeners?

[zacknewman]

Before I make my own crate, I wanted to ask if it makes sense to add "dual-stack" listeners to tokio::net? Right now only certain platforms can use a single TcpListener/UdpListener to handle both IPv6 and IPv4 requests; furthermore most of those platforms are still dependent on OS configuration (e.g., the default value of IPV6_V6ONLY) and has the unfortunate requirement of binding to the "wildcard" IPv6 address which in turn prevents other services from binding to the same TCP/UDP port but different IP.

I've written a few applications now based on copy-and-pasted code I've written before that adds proper OS-independent dual-stack listeners. I've done this enough that I would like to either submit a PR that adds such ability to tokio proper or at least publish my own crate. The types would basically be newtypes around two TcpListeners/UdpListeners that defer method calls like accept to the contained listeners. They'd do so in a way that prevents one IP version from "starving" another by alternating which listener gets polled first. Since the listeners are "cancel safe" per the docs, it's perfectly fine to yield the appropriate stream once one does (essentially accept would be based on code that's roughly equivalent to what select! does). A function like bind, however, would not yield until both listeners yield (i.e., roughly equivalent to what join! does). Anyway, I thought I'd ask.

[Darksonn]

No, that sounds like it would be out of scope.