In article <20050714030200.ABC5A63B121@narn.netbsd.org>,
	Martin Husemann <martin@duskware.de> writes:
> 
> When upgrading a machine to 3.0_BETA or -current, ntpd might stop working.
> If you have a time server in your /etc/ntp.conf that has a IPv6 address,
> like the former public server at ntp.univ-lyon1.fr:
> 
> ntp.univ-lyon1.fr is an alias for dns.univ-lyon1.fr.
> dns.univ-lyon1.fr has address 134.214.100.6
> ntp.univ-lyon1.fr is an alias for dns.univ-lyon1.fr.
> dns.univ-lyon1.fr has IPv6 address 2001:660:5001:100::6
> ntp.univ-lyon1.fr is an alias for dns.univ-lyon1.fr.
> 
> and you are running a IPv4 only kernel, ntpd logs:
> 
> sendto(2001:660:5001:100::6): Address family not supported by protocol family
> 
> and dies!
> 
> I'm sure this is part of the conspiracy to force everyone into IPv6 - but
> please start pushing ISPs first instead of users.

Ha! Indeed. I'm starting to think that global IPv6 is never going
to happen.

You can put "-4 " before the name of the server to force IPv4.
Does that help?

There's an open request to have ntpd try all addresses in turn:

  https://ntp.isc.org/bugs/show_bug.cgi?id=334

A more general solution would be to train the resolver to do IPv4
only.  In lieu of that, if you can run a nameserver, the latest
version of bind9 has a "-4" option, which is "politically correct"
by virtue of there also being a "-6" option.

-- 
Frederick