feat: support ipv6 link-local address

[jclab-joseph]

Close #2659 .
Make it possible to connect to the link-local address.

Note that
To work with mdns, must also modify https://github.com/libp2p/zeroconf.

[marten-seemann]

Is there any reason why a well-behaved peer would ever advertise an ip6zone address? I can’t see any, but I might be missing something.

If that’s the case, we don’t need to make any changes to the TCP transport.

[jclab-joseph]

@marten-seemann

As an example, an IoT node before IP is set can use it to communicate via p2p.
IPv6 link-local address can be used even if a static IP is not set.
In this case, you can communicate by searching for each other using mdns, etc. within the local network.

Since link-local address cannot be used on the Internet, it was intentionally excluded from announce.

[marten-seemann]

What do you mean by „before the IP address is set“? Are you referring to DHCP?

How would and ip6zone address passed to the TCP transport look like?

[jclab-joseph]

@marten-seemann
To be exact, it is an environment without DHCP. In an environment with DHCP, there is no need to use link-local address.
e.g. /ip6zone/eth0/ip6/fe80::3811:2aff:fe94:adbe/tcp/1234

[marten-seemann]

Can you just advertise /ip6/fe80::3811:2aff:fe94:adbe/tcp/1234? For the peer, it’s irrelevant that it’s on your eth0 interface, right?

[jclab-joseph]

@marten-seemann Currently, link-local addresses are intentionally excluded from basic_host.
Is there a reason to announce it? A link-local address is only useful if the receiving party can determine the listening interface, such as mdns.
I thought it would be a good idea not to advertise unnecessary addresses. Still, is it better to advertise?

[marten-seemann]

A link-local address is only useful if the receiving party can determine the listening interface, such as mdns.

I don't understand how this would work. What is configured to be eth0 is only relevant to me, not to the remote node. Therefore, we should never advertise it.

[marten-seemann]

Please don't reply here. Let's move discussion of this back to the issue: #2659.

[marten-seemann]

I'm adding the blocked label here, since there's little point in moving this PR forward until we've resolved libp2p/zeroconf#35.

Once that has been done, we still need to decide if support IPv6 link-local is something we want to do, see #2659 (comment).

[MarcoPolo]

Automatically closing old open PRs. If you're still interested in merging this please open the PR again and rebase from the latest master, otherwise do nothing.