Prefer IPV6

[amarokuser]

I'd like to prefer ipv6 peers over ipv4. As long as there are ipv6 peers to connect to, no attemt is made to connect a ipv4 peer. If there is no ipv6 peer left to connect to than continue with ipv4 peers.

[jch]

I'd like to prefer ipv6 peers over ipv4. As long as there are ipv6 peers to connect to, no attemt is made to connect a ipv4 peer. If there is no ipv6 peer left to connect to than continue with ipv4 peers.

Why?

--jch

[Rolcol]

Guys, jch is the DHT and IPv6 master.

[amarokuser]

Seeing some interest I filed a patch of my idea this could be made. Actually it is a little bit more that might be of public interest too.

[jch]

Because peers who make the effort to use a new technology should be rewarded with connections.

They already are, indirectly. A peer that is participating in both v4 and v6 is reachable for both v4-only and v6-only peers. (And while there aren't many v6-only peers, the few ones that are there tend to be fast peers on University networks.)

I'm not sure to what extent it's a good idea to further this effect, especially since I'm always worried about swarms getting partitioned.

--jch

[livings124]

Preferring peers based on factors besides ability to upload/download (and similar) seems like a bad idea. Definitely would lead to hurting the swarm as a whole.

[jch]

since I'm always worried about swarms getting partitioned.

Thinking about it some more, the following scenario is not all that unlikely.

Imagine a swarm with 120 peers, 60 of which are v4-only peers, and 60 of which are double-stack (v4+v6) peers. If the double-stack peers favour v6 connections, then you might end up with the double-stack peers all connecting to each other, and not accepting any connections from the v4-only peers, which will then only connect to each other. You end up with a partitioned swarm.

There are probably enough other factors in BitTorrent to ensure the swarm ends up recovering, but it still doesn't seem like a good idea.

--jch

[amarokuser]

Thinking about it some more, the following scenario is not all that unlikely.

According to my observations it is almost unlikely to happen. It goes like this: Native IPV6 is rare. Most of IPV6 peers are 6to4 or Teredo. These peers seem to get superseded by better performing IPV4 peers after a while.

since I'm always worried about swarms getting partitioned.

IPV6 is meant to succeed IPV4 so it is unavoidable to "partition" the swarm.

My intention is to get more (bulk) traffic to IPV6.

[User294]

In theory, nothing prevents all molecules of the air to leave your room just to fill another one. This event is possible but very unlikely to happen. So we do not need to wear protective suites to be able to survive in vacuum. Same goes for other probabilistic processes as well. The probability that IPv4 peer would fail to see any other IPv4 capable peer at all would be almost zero for most of swarms. And at least theoretically, IPv6 causes less load on routers so it could be somewhat faster, so preferring it and encouraging it's use could be not a bad idea, actually.