Comment 1 by bortzmeyer:

Regarding the question of "is net.ipv6.bindv6only=1 broken?", Java people have a
similar issue: 
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6342561

Comment 2 by bortzmeyer:

Rather than arguing if net.ipv6.bindv6only=1 is broken or not, a better solution may
be to use the socket option IPV6_V6ONLY (RFC 3493, section 5.3).
This option allows a per-socket tuning of the "V6only binding". Go could set it to 0
systematically or, better, export it for applications to use.
Apparently, with today's tip, this option is imported
(src/pkg/syscall/zerrors_linux_386.go) but I'm not sure it can be used by
applications. (syscall.SetsockoptInt ? Not very clean)

Comment 3:

I'm going to fix it by using that option always.
That's why it's there.
Just haven't cleaned up and submitted the code yet.

Comment 4:

Issue #685 has been merged into this issue.

Comment 5:

Issue #890 has been merged into this issue.

Comment 6 by joerg.sonnenberger:

Attached is a patch that works on NetBSD.

Attachments:

1. v6only.diff (3209 bytes)

Comment 7:

This issue was closed by revision d472665.

Status changed to Fixed.