Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 111 Next »

If you know of any issues that customers might encounter during the transition to IPv6, please add a subheading to this page and explain both the problem and how to address it.

«Broken» users unable to access dual-stacked content

A small number of users have some kind of misconfiguration or bug in their internet connection that makes them unable to properly access dual-stacked web sites. More often than not, these users have no problems accessing IPv4-only sites, which causes them to perceive the dual-stacked site as the one having problems. A web site operator can detect these users using JavaScript and potentially warn them about the problem, see Warning broken users with JavaScript.

The existence of these users is unfortunately causing content providers to put off enabling IPv6. For more information, see these presentations by Yahoo ( Fesler Y\!atGIPv6ImpConf.pdf?attredirects=0), Google, and Redpill Linpro. There's also an article about the problem in Wikipedia.

This section attempts to document the most common causes as to why this happens, and how end users can solve it. It is based on real operational experience from running dual-stacked web sites.

Use of transitional IPv6 connectivity

Transitional IPv6 connectivity (6to4 or Teredo), is usually less reliable than IPv4. For this reason, the end user's web browser should prefer to use IPv4 when attempting to access dual-stacked site. This is well documented in I-D.vandevelde-v6ops-harmful-tunnels, as well as by researchers Emile Aben and Geoff Huston.

6to4 router functionality is implemented and often enabled by default in a large number of home gateway products made by several different vendors. This is unfortunately in accordance with Microsoft's requirements for Windows-compatible home routers. Due to the fact that not all operating systems and applications de-prefer transitional IPv6 connectivity, it is recommended to disable it (especially 6to4) whenever possible.


Earlier versions of Android preferred transitional IPv6 connectivity above IPv4, as its system resolver did not implement RFC 3484.

Solution: Upgrade Android to version 2.2 Froyo or later.

Apple Mac OS X

Mac OS X, in versions earlier than 10.6.5, do not de-prioritize 6to4 compared to IPv4.

Solution: Upgrade Mac OS X to the latest available version.

Versions predating 10.6.x «Snow Leopard»

At the time of writing, the patch that de-prioritizes 6to4 is only available for the «Snow Leopard» series (10.6.x). Avoiding 6to4 entirely is the best solution, but might not always be possible or feasible when another device in the network is announcing itself as a 6to4 router.
Users on old PowerPC systems have no upgrade path from 10.5.

Work-around: Disable IPv6 completely in the operating system: System Preferences -> Network -> Advanced -> TCP/IP -> Configure IPv6 -> Off.


The GNU C Library will de-prefer transitional IPv6 connectivity if the local IPv4 address is not a private one (RFC 1918), due to a strict interpretation of RFC 3484 (see SW#11438).

Solution: Upgrade the Linux distribution to Debian Squeeze, Fedora 13 Goddard, Ubuntu 10.04 Lucid Lynx, Mandriva 2010.1 Spring, openSUSE 11.3, Gentoo 2010-04-25, or any later versions.

Alternative solution: Add the following lines to the file /etc/gai.conf (create it if it doesn't exist):

Microsoft Windows

MS Windows automatically enables Teredo and 6to4 whenever possible. The system resolver library will de-prioritize their use, but not all applications use the system resolver library and will therefore end up preferring the less reliable transitional IPv6 connectivity.

Solution: Disable 6to4 and Teredo, by entering the following commands in an Administrator shell:

Internet Connection Sharing

If Internet Connection Sharing is enabled on any interface (it doesn't even have to be connected), the Windows hosts will announce itself as an IPv6 router to the local network. This will in turn cause problems for other operating systems that doesn't de-prioritize transitional IPv6 connectivity, notably Mac OS X (see above).

Solution: Disable 6to4 and Teredo as described above, and also ensure Internet Connection Sharing is disabled whenever unused.

Opera web browser

The Opera web browser uses its own resolver library, and would in versions earlier than 10.50 (on Microsoft Windows) and 10.63 (on Mac OS X and Linux) prefer transitional IPv6 connectivity above IPv4.

Solution: Upgrade Opera to the latest available version.

Home gateways and broadband routers


The AirPort and Time Capsule wireless base stations will advertise the prefix ::/64 on their LAN interfaces if they are set up with IPv6 mode tunnel while having a private (RFC 1918) address assigned to their WAN interfaces. As the resulting auto-configured addresses are invalid, hosts on the LAN will suffer dual-stack breakage. The bug was last reported in firmware version 7.4.2, and is confirmed to be fixed in firmware version 7.5.2).

Solution: Upgrade the firmware of the AirPort/Time Capsule to the latest available version (instructions).


Certain AVM FRITZ!Box models (at least 7270 International v2) is known to have ULA (RFC 4193) functionality based on recommendations from I-D.ietf-v6ops-ipv6-cpe-router revision 07 or earlier. When enabled, this leads to dual-stack problems for all hosts behind the router due to the fact that ULA addresses aren't special-cased by any modern operating system. ULA addresses cannot be used for connectivity to the global IPv6 internet.

Solution: Ensure ULA functionality is disabled. This can be done from the router's web interface - instructions follows:

  1. Open with your web browser.
  2. Enter your FRITZ!Box password.
  3. Click Settings (towards the top of the window).
  4. Click Internet.
  5. On the left, click Account Information.
  6. Select the IPv6 tab on the right side, towards the top of the window, under the heading Account information.
  7. You are now on the IPv6 settings page. The last section of this page is called Unique Local Addresses. In this section, select do not assign unique local addresses (ULA) (not recommended).
  8. Click Apply.

Cisco Linksys

E2000, E3000, and E4200

These models (and possibly others which share the same code) offers by default DHCPv6 service that assigns addresses to the LAN hosts from the documentation prefix 2001:db8::/32. However, no ICMPv6 Router Advertisements are being sent in this case, which causes the LAN hosts to not have a default route, thus causing the host operating stack to generate fast internal failures when IPv6 is attempted.

They also enable 6to4 router functionality by default. If 6to4 is active (i.e. if the WAN interface is numbered using a public IPv4 address), ICMPv6 Router Advertisements will be transmitted to the LAN segment, containing a Prefix Information Option for the 6to4 prefix.

Luckily, the networking stack in Microsoft Windows will immediately remove the DHCPv6-assigned bogon address from 2001:db8::/32 upon receiving an ICMPv6 Router Avertisement. As Microsoft Windows is the only major operating system that supports DHCPv6 (and uses it even when there are no ICMPv6 RAs), end users are unlikely to experience brokenness due to this issue.

The recent firmware revisions disable both 6to4 and the DHCPv6 service with the bogon addresses by default, solving all known problems.

Solution: Upgrade the router firmware to the firmware version indicated below (or newer):

Select the Downloads tab on the pages linked to above in order to acquire the latest firmware image and its corresponding release notes.

IPv6/6to4 settings can be configured from (assuming is the IPv4 address of the device):


This model (and possibly others) uses the bogon range 2005:123:456:789::/64 as its default DHCPv6 pool. These addresses cannot be used to communicate with the IPv6 internet, and their presence on end-user hosts will cause end-user brokenness.

Solution: Ensure the router is operating in IPv4-only mode. This setting is found under Setup -> IP Versions in its web interface.

Several D-Link models from the DSL series (at least DSL-584T, DSL-G624T, DSL-G664T, and DSL-G684T), do not correctly forward DNS responses for hostnames with both A and AAAA records published. What it does is to stuff the first 32 bits of the AAAA record into the A record that's being returned to the end user's computer. In other words, will incorrectly resolve to (the 2001:0500: part of the IPv6 address). If the operating system or web browser prefers to use IPv4, it will be unable to connect to the destination.

Italian ISP Wind/Infostrada is reported to have distributed the DSL-G624T to its customer base over a period of several years.

It doesn't happen all the time - it appears to be timing-dependent. Older Mozilla Firefox browsers are hit particularly bad, due to the fact that they will request AAAA lookups even if the local host does not have an IPv6 address.

Solution: Attempt to upgrade the firmware of the router to the latest version. If that doesn't help, replace the router.

Solution/work-around: If the user is using Mozilla Firefox, upgrade to version 4.0 or newer.

Alternative work-around: Disable IPv6 in the operating system, which will in most cases suppress AAAA lookups.

Bugs in operating system TCP/IP stacks

Apple iOS

No fallback from IPv6 to IPv4

Apple iOS is not able able to fall back from IPv6 to IPv4, if the initial IPv6 connection attempt fails due to blackholing. An error message is displayed after a timeout of about 60 seconds instead. If errors are generated by the network, on the other hand, it will successfully fail over; either instantly (TCP RSTs), or after four seconds (ICMPv6 unreachables).

Short of ensuring that any IPv6 connectivity present work perfectly, there is no known workaround. IPv6 cannot be disabled in iOS.

This bug is reported to Apple in #8702877. The latest iOS version the bug has been confirmed to be present in is 4.3.

Apple Mac OS X

Use of IPv6 with only link/site-local addresses

When attempting to connect to a dual-stacked destination when the system has only link-local IPv6 addresses (but at the same time a default IPv6 route), a 75 second timeout is incurred per AAAA record. IPv6 is preferred over IPv4 in this case due to lack of support for RFC 3484 in Mac OS X's system resolver. A system will be vulnerable to this bug if it has received an ICMPv6 Router Advertisement message that does not contain any Prefix Information Options, for example.

Portugese ISP SAPO is known to distribute a CPE to its customers (Pirelli A1000G) that emits such RAs by default, and the D-Link DIR series of routers will do so as well (at least DIR-655, DIR-825, and DIR-855).

Hosts running Microsoft Windows (at least Vista) with Internet Connection Sharing enabled can also emit such prefix-less RAs, in certain situations (possibly limited to when two network interfaces are bridged).

This will especially affect users of Mozilla Firefox older than version 4.0, as it will request AAAA records even though the system does not have any global IPv6 addresses (see bug #614526), and will attempt to connect to the IPv6 addresses in preference to falling back to IPv4. Safari also suffered from this problem up until Mac OS X 10.6.4. It will also affect users of the virtualisation software Parallels Desktop (regardless of the browser used), because its virtual network interfaces have site-local IPv6 addresses assigned, in turn making the AI_ADDRCONFIG flag to getaddrinfo() ineffective.

Solution/work-around: If the user is using Mozilla Firefox, upgrade to version 4.0 or newer.

Work-around: Disable IPv6 completely in the operating system: System Preferences -> Network -> Advanced -> TCP/IP -> Configure IPv6 -> Off.

Similarly, the "ssh" client application (accessible in a MacOS X shell) initiates connections to global scope IPv6 addresses even if the OS only has link-local addresses in its interfaces.

Solution: None. Bug with ID 8820407 is open with Apple.

Handling of Router Advertisements with a lifetime of 0

When receiving an Router Advertisement packet with a lifetime of 0, Mac OS X-based hosts will incorrectly install a default route. It appears that a Prefix Information Option must be present in the RA packet for the bug to take effect.

The technique of using a lifetime of 0 to announce an IPv6 router without global connectivity is used by devices conforming to I-D.ietf-v6ops-ipv6-cpe-router in the case where ULA addressing are used within the residental site. In this situation the Mac OS X host will attempt to use ULA IPv6 addresses for global connectivity, leading to timeouts towards dual-stacked content.

Apple bug ID: 8705091.

Work-around: Disable IPv6 completely in the operating system: System Preferences -> Network -> Advanced -> TCP/IP -> Configure IPv6 -> Off.

Handling of ICMPv6 Destination Unreachable

When a router in the network responds to a TCP SYN packet with an ICMPv6 Destination Unreachable, Mac OS X will re-transmit the TCP SYN packet five times with intervals of one second before giving up, even though the ICMPv6 code indicates that the situation is permanent and further attempts are pointless (e.g. Code 2 - Beyond scope of source address). In other words, an avoidable four-second timeout is incurred for every outgoing TCP connection.

Microsoft Windows

Handling of ICMPv6 Destination Unreachable

Microsoft Windows appears to ignore received ICMPv6 Destination Unreachable completely, instead falling back on the overall connection timeout mechanism. This causes a completely avoidable 21 second long timeout to occur for every outbound TCP connection.

Bugs in web browsers

Mozilla Firefox

Mozilla Firefox does not set the AI_ADDRCONFIG flag when looking up names using the system getaddrinfo() library function, which causes it to solicit AAAA records even though the system has no IPv6 addresses (non-loopback and non-linklocal). This can trigger other problems, such as the D-Link AAAA mangling bug and the Mac OS X bug regarding the use of link-local IPv6 addresses when connecting to global destinations. The bug is fixed as of Firefox 4.0, see the bug report

Solution/work-around: Upgrade to Firefox version 4.0.


When Opera, in versions older than 11.10, is used on a Mac OS X machine that has Parallels Desktop installed, it is unable to connect to dual-stacked web sites (except if the machine in question also has connectivity to the IPv6 internet). The bug is reported to Opera Software; ID is DSK-326913.

Solution: Upgrade Opera to version 11.10 or later (currently only available as a beta release)

Work-around: Enable Opera Turbo (the setting is available in the Quick Preferences menu).

If the use of pre-release software, or Opera Turbo, is not desired by the user, either un-install Parallels Desktop (using the recommended method), or use another web browser.

Firewall Config Issue

Especially if you have users on Vista.
It does this IPv6 tunnelling thing that on the surface
appears really cool. When you try and talk IPv6 to
something other than link-local: (in order)

  • If you have a non-RFC1918 (ie. 'public') address, it fires up 6to4.
  • If you have an RFC1918 address, it fires up Teredo.
    Seems cool in theory, and you'd think that it would really help global IPv6 deployment - I'm sure that's how it was intended, and I applaud MS for taking a first step. But in practice, however, this has essentially halted any IPv6 /content/ deployment that people want to do, as user experience is destroyed.

You can help, though - here's the problem:

6to4 uses protocol 41 over IP. This doesn't go through NAT, or stateful firewalls (generally). Much like GRE.
Because of this, if you're a enterprise-esque network operator who runs
non-RFC1918 addresses internally and do NAT, or you do stateful firewalling, PLEASE, run a 6to4 relay on internally, but return
ICMPv6 unreachable/admin denied/whatever to anything that tries to send data out through it. Better yet, tell your firewall vendor to allow you to inspect the contents of 6to4 packets, and optionally run your own 6to4 relay, so outgoing traffic is fast.
Even if you don't want to deploy IPv6 for some time, do this at the very least RIGHT NOW, or you're preventing those of us who want to deploy AAAA records alongside our A records from doing so. If you need configs for <vendor/OS B/C/J/L>, post a message to the NANOG list and I'll write some templates.

I see this sort of IPv4 network quite commonly at universities, where students take their personal laptops and throw them on the campus 802.11 network. While disabling the various IPv6 things in Vista at an enterprise policy level might work for some networks, it doesn't for for a university with many external machines visiting. So, if you're a university with a network like this (ie. most universities here in NZ, for example), please spend a day or two to fix this problem in your network - or better yet, do a full IPv6 deployment.
--Nathan Ward

Increased Latency to your IPv6 Content

If you do deploy an IPv6 network for your content, set up a Teredo relay, and point 2001::/32 at it. Your viewers/users will automatically use this relay when accessing your content, and their traffic to you will be over IPv4, all they way from their PC to your network - so, equivalent performance as IPv4. Note that I say relay here, not server. are doing this for example. Cue Matthew Zeier.
--Nathan Ward

Check out Enabling IPv6 on a Mail Server‎.

Unfortunately, not all Domain Registrars are providing IPv6 Glue yet

It may be tough to get IPv6 AAAA records for your nameservers into the DNS Glue records, depending on your registrar.

Check out DNS Registrars IPv6 Support Status, let's build a list of who does and does not.

  • No labels