IPv6 support shouldn't be Optional

[xlmnxp]

I'm using IPv6 for all of my internet stuffs and I face some sites/applications that not support IPv6 because they use tools that not support IPv6 or harder for them to support it

me and many users want to use IPv6 and adoption of it more then 40% in the world and it will the present that must support

[xlmnxp]

not support Dual-stack mean it will be harder for developers ( which many of them lazy ) to support IPv6

[badeend]

I believe there are a couple of related, but separate issues here:

---

IPv6 support shouldn't be optional

I assume you specifically mean that the create-tcp/udp-socket WASI functions should not fail just because ip-address-family::ipv6 was passed. Seeing that pretty much every OS supports IPv6, I don't think there is any practical reason for concern.

---

not support Dual-stack mean it will be harder for developers ( which many of them lazy ) to support IPv6

Regarding dual-stack support, please refer to #1 .

---

me and many users want to use IPv6 and adoption of it more then 40% in the world

I would love to see greater adoption too. But I think most of the trouble comes from: adapting existing (user space) applications and all the middle boxes on the network. Besides allowing IPv6 sockets to be created from WASM, there's not much else we can do at the WASI level.

[rwally1986]

Supporting Dual-Stack (IPv4 preferred, fail over to IPv6, IPv6 preferred, fail over to IPv4) provides resilience for the developers application, resulting in wider choice of network transits and resulting in better user experience. In short, WASM should support both IPv4 and IPv6 during the move from 40% till ~90-99% (4-12 years), as IPv4 is deprecated and removed from carriers and hosting/cloud providers.

[xlmnxp]

Supporting Dual-Stack (IPv4 preferred, fail over to IPv6, IPv6 preferred, fail over to IPv4) provides resilience for the developers application, resulting in wider choice of network transits and resulting in better user experience. In short, WASM should support both IPv4 and IPv6 during the move from 40% till ~90-99% (4-12 years), as IPv4 is deprecated and removed from carriers and hosting/cloud providers.

I think IPv6 must preferred then IPv4 when fail like what is standard in browsers and operating systems
http://test-ipv6.com/faq_avoids_ipv6.html

[badeend]

Do you think there is something that needs to be changed to the wasi-sockets proposal?

[PHJArea217]

wasi-sockets, like all wasi interfaces, is only an interface. It does not place constraints on the "black box" behavior of the implementation, only that which is visible to the application. Which means that it's totally valid for a WASI runtime for the IPv4 socket APIs to use IPv6 as underlying network transport. Or they could be sent via a http or socks proxy. Or routed through a Unix domain socket or named pipe. And all of those would theoretically still be valid for a WASI implementation.

(I came here because I recognized @xlmnxp here.)