Chris can you share details of the brokenness check? What variables are considered?
On Thu, Oct 10, 2013 at 12:02 AM, Christopher Palmer < [email protected]> wrote: > John and Lorenzo beat me to it J.**** > > ** ** > > Example:**** > > Samantha has native IPv6 and Teredo.**** > > Albert has Teredo only.**** > > ** ** > > Albert, in destination address selection, will chose Samantha’s Teredo > address. Samantha, in source address selection, will use her Teredo > address. This will avoid relay traversal.**** > > ** ** > > Xbox P2P policy is a bit more sophisticated than RFC 6724, but I note that > the avoidance of Teredo relays is also part of Windows behavior. Windows > address selection is a fairly clean implementation of RFC 6724. In RFC 6724 > terms, Teredo -> Teredo is a label match (Rule 5), Teredo -> Native IPv6 is > not. The biggest difference between us and the standard is the brokenness > check.**** > > **** > > This does complicate the dream. In order for a set of peers to use native > IPv6 – BOTH peers have to have native available. In the pathological case, > if half of the world has IPv6 and connects only to the other half that only > has Teredo, and no one actually uses native IPv6.**** > > ** ** > > Realistically, matchmaking is going to prefer users “close to you” (and a > bunch of other things, like their gamer behavior and stuff). Naively I > expect IPv6 traffic to start as local pockets, Albert playing against his > neighbor, both with the same ISP. As IPv6 penetration grows hopefully we’ll > see significant P2P traffic across the Internet use native IPv6 transport. > **** > > ** ** > > ** ** > > *From:* > ipv6-ops-bounces+christopher.palmer=microsoft....@lists.cluenet.de[mailto: > ipv6-ops-bounces+christopher.palmer=microsoft....@lists.cluenet.de] *On > Behalf Of *Lorenzo Colitti > *Sent:* Wednesday, October 9, 2013 8:26 PM > *To:* Geoff Huston > *Cc:* IPv6 Ops list; Christopher Palmer > > *Subject:* Re: Microsoft: Give Xbox One users IPv6 connectivity**** > > ** ** > > On Thu, Oct 10, 2013 at 12:19 PM, Geoff Huston <[email protected]> wrote:**** > > But I've thought about your response, and if I'm allowed to dream (!), and > in that dream where the efforts of COmcast, Google etc with IPv6 bear > fruit, and I'm allowed to contemplate a world of, say, 33% IPv6 and 66% V4, > then wouldn't we then see the remaining Teredo folk having 33% of their > peer sessions head into Teredo relays to get to those 33% who are using > unicast IPv6? And wouldn't that require these Teredo relays that we all > know have been such a performance headache?**** > > ** ** > > Can't you fix that by telling the app "if all you have is Teredo, prefer > Teredo even if the peer has native IPv6 as well"?**** > > ** ** > > Of course this breaks down when IPv4 goes away, once IPv4 starts going > away then there's really way to do peer-to-peer without relays, right? > (Also, IPv4 going away is relatively far away at this point.)**** >
