Talk:Teredo tunneling

Triangle router use appears wrong

> and a costly relay would have to be used to perform triangle routing.

My understanding of triangle routing has mobile device A sending to B directly but with a care/of return address of a mobile proxy that knows where A is now. C forwards the message. A is moving and it's IP changes as it crosses cells. Thus B sends care/of C which forwards to A. A continues to send directly to B. This forms a triangle A->B->C->A.

I think they intended to talk about an outbound proxy that allows symmetric nat traversal. With an outbound proxy neither A or B communicate directly but both send packets to the outbound proxy which forwards all communication. This forms a V. A->D->B->D->A

I am still developing an understanding of this stuff and so did not want to make a change without validation. Although linked there is no triangle routing page. —Preceding unsigned comment added by 69.157.5.200 (talk) 20:12, 29 November 2008 (UTC)

Triangle routing in this context refers to all the packets between A and B doing a detour via an overloaded central server C. Anyway, RFC 6081 now provides some workarounds for using Teredo through various kinds of symmetric NAT. 77.215.46.17 (talk) 01:57, 26 November 2012 (UTC)

Teredo relay route export

I could find no evidence of any BGP export requirement for Teredo relays in RFC 4380. For the Teredo service to operate properly, at least one relay as to export a route through BGP, because not all autonoumous systems will ever have their own Teredo relays. That does not imply all Teredo relays have to do BGP export (the same considerations apply to 6to4 relays).

RFC 4380 actually refers to “the IPv6 routing protocols” (§ 5.4, emphasis added), which supposedly include not only BGP, but also interior routing protocols (OSPF, IS-IS, or even RIPng and manual setup) and adds that a relay might serve as few as a single IPv6 host, i.e. itself.

In addition to that, RFC 4380 also mentions the “range of IPv6 addresses served by the relay” (§ 5.4.3). This hardly make sense if all relay have to use BGP, in which case the range of IPv6 would supposedly be the whole IPv6 Internet.

This accounts for my reversal of Jec's commit 54646349 (03:34, 23 May 2006). Someone might still want to clarify the text involved.

-- Rdenis

Confusion on my end -- the “direct connectivity test” procedure makes it possible to have symmetric routing without global IPv6 reachability of all relays. Thanks for correcting. --Jec 19:39, 28 May 2006 (UTC)

Temporary Measure

If this takes off in a big way (and it probably will) then this won't be a 'temporary measure'. It will be around as long as people/companies can use it instead of replacing old routers etc. —Preceding unsigned comment added by 217.20.17.18 (talk) 15:02, 14 August 2008 (UTC)

Most probably not, since native support is, well, native, and much easier than using external programs and slower-than-required paths. This method seem to introduce arbitrary delays on the path, which make it less ideal to use as native. --grin ✎ 09:44, 28 January 2010 (UTC)

Also, some of the current Teredo Relays out there are getting close to a gigabit of traffic already, which is already painful to route on cheap PC hardware. Many of them will opt to reduce visibility of their router or stop providing the service altogether rather than invest in bigger hardware. Teredo cannot be easily scaled by just adding more PCs. This will of course add more load on the remaining relays, which will make them also want to quit the job. -- Aleksi Suhonen, January 2011

Fix bad wording

I cannot reliably fix without background knowledge and I don't want to mess it, so someone please replace the words "host" with the proper phrase (client, relay, server, ...) in the section Teredo relays (and maybe other sections, I'm halfway through reading) in many places. --grin ✎ 09:42, 28 January 2010 (UTC)

Citation at www.atm.tut.fi

I work at the Department of Communications Engineering at Tampere University of Technology. www.atm.tut.fi is one of our old websites, from the 90s when ATM was hip and cool. Since then the department has had many face lifts and as a result, most of our old websites will probably be discontinued soon. I actually thought that www.atm.tut.fi had already been discontinued...

Anyway, the citation refers to an email message by Huitema on the ngtrans mailing list hosted by Sun Microsystems. The said mailing list probably has more official archives than the one the citation uses. Google found me the following link which seems as official an archive for the mailing list as is possible:

http://playground.sun.com/pub/ngtrans/mail-archive/ngtrans.200201

Unfortunately it's in mbox format where the www.atm.tut.fi archive is pretty printed and split into individual messages. So, rather than switching the citation link myself to this uglier one, I thought I'd ask the community to decide....

--Aleksi Suhonen, January 2011 —Preceding unsigned comment added by 91.155.191.20 (talk) 01:58, 17 January 2011 (UTC)

Switched to IETF archive, it is uglier, but better than a broken link. --Cybjit (talk) 19:22, 17 January 2011 (UTC)

questions

i cannot comprehend this article.

from what i gather, a teredo relay is like a proxy or "middle man" that an ipv4-only connected device can send packets to, and then that relay will convert the packet into an ipv6 packet and send it to its destination.

so if microsoft have enabled teredo on its newer operating systems, would that not mean huge amounts of bandwidth bills being charged to microsoft? and why does no one seem to be concerned about all their teredo traffic being routed through microsoft-owned servers (by default)?

one other thing, why can a computer set up to use teredo not connect to ipv6.google.com or other ipv6 sites? is it because both communicating devices must understand the protocol? —Preceding unsigned comment added by 76.182.100.79 (talk) 00:21, 18 April 2011 (UTC)

Lots of people run teredo relays, not just Microsoft, so the pain is shared around.Rememberway (talk) 00:52, 18 April 2011 (UTC)

The other thing is that you can get to ipv6.google.com using teredo (I've done it), after the packets go through a teredo relay they can then be routed through ipv6 in the normal way, although some teredo relays won't give you a route there, the normal ipv6 routing will automagically find one that will.Rememberway (talk) 00:52, 18 April 2011 (UTC)

Afaict for teredo to work the following have to hold

1: the ping from the teredo server must reach the ipv6 host

2: the ipv6 host must respond to ipv6 ping

3: the ping response must reach a teredo relay

4: the teredo relay must successfully return the ping response to the client (which involves NAT traversal techniques)

5: the client must send the relay the first real packet (if the above worked this should be pretty much a certainty)

6: the teredo relay must successfully send the first real packet to the ipv6 host.

In short there are LOTS of things that can go wrong with a teredo connection, not to mention some hosts may delibeately block it for whatever reasons (legitimate or otherwise). 80.0.68.41 (talk) 13:22, 2 May 2011 (UTC)

Disambiguation/Redirect: Shipworm or Teredolites

I am surprised that "Teredo tunnel" doesnt direct to the article on Shipworms. Shouldn't there be at least a link to the disambiguation page on top, or is the paleontological meaning of the word not used in English? --Polymeris (talk) 08:46, 18 April 2011 (UTC)

you only do that if it's likely, but it's unlikely that somebody would get to the teredo tunneling article when they were looking for shipworms. If they type in teredo on its own they get the disamb page.Rememberway (talk) 14:11, 18 April 2011 (UTC)

Happened to me. (Not saying that makes it likely) I wasn't looking for shipworms in general, but Teredolites a.k.a. "Teredo tunnels", which I thought was a (relatively) common term. --Polymeris (talk) 18:34, 18 April 2011 (UTC)

Example potentially confusing

Does anyone else feel giving an example where the public IPv4 is in the space reserved for private LAN and therefore commonly the LAN IPv4 of clients behind NAT potentially confusing? I understand we wouldn't want to use some random IPv4 of some random ISPs client, but how about something else like a Microsoft or Google or even wikimedia IPv4 (yes these may change but it's no big deal) 2001:0:4137:9E76:247C:A71:833A:FA41 (talk) 14:07, 7 June 2012 (UTC)

The example 192.0.2.45 is not in the reserved spaces for local addresses (10.x.x.x, 172.16.x.x-172.31.x.x, 192.168.x.x and 169.254.x.x) but in the address space reserved for examples, documentation and fiction (192.0.2.x). 192.0.2.x addresses are made up public IP addresses that exist only in made up examples, just like North American 1-555-555-01xx phone numbers and Internet domains that contain the word example as last or second last part (silly.example, silly.example.com, silly.example.org etc.). 77.215.46.17 (talk) 02:19, 26 November 2012 (UTC)