6to4 route at AS27552? (AS7922?)

Ivan Shmakov oneingray at gmail.com
Sat Nov 10 22:18:56 CET 2012


>>>>> Gert Doering <gert at space.net> writes:
>>>>> On Fri, Nov 02, 2012 at 12:01:23PM +0700, Ivan Shmakov wrote:

[…]

 >> I wonder what would be the cost of disabling 6to4 on each and every
 >> single networking device willing to connect 2001:4830::/32 (and has
 >> 6to4 currently enabled), as compared to the cost of fixing one
 >> single route?

	JFTR: the issue was resolved around 2012-11-09 21:06 UTC.
	Thanks to those involved.

 > You don't need to specifically disable 6to4, if those devices are
 > well-behaved and prefer IPv4 to 6to4,

	Somehow, I'm not aware of a free software resolver library
	that'd do that.

	Also, somewhat surprisingly, my GNU/Linux hosts seem to prefer
	6to4 IPv6 addresses over non-6to4 ones (where there're both)
	when initiating connections.  (It may be some misconfiguration
	on my part, though.)

 > and you're not connecting to IPv6-only services from
 > IPv4-only-plus-6to4 clients, and you're not putting 2002:: addresses
 > as destinations into DNS.

 > "Fixing 6to4 with anycast relays run by volunteering networks that do
 > not guarantee anything" is not possible.

 > Emphasis on the "anycast relays" bit.

	Which makes me wonder on what are the costs of operating one's
	own “IPv6-to-6to4” relay?  As it seems, the “no valid route to
	192.88.99.1” case is much easier to troubleshoot that the
	converse “no valid route to 2002::/16” one, so the latter may
	indeed deserve some extra care.

-- 
FSF associate member #7257




More information about the ipv6-ops mailing list