Ghost bit-torrent tracker

January 12, 2008 7:50 AM

How do torrents work?

I thought I knew. Something like this, basically:

Person A wants to share files. She creates a .torrent which contains
file-list +
pieces & their checksums +
tracker address +
maybe some other stuff

She uploads it to tracker, which loads the torrent, and if there's an associated directory, also gets it indexed.

Persons B, C, D come across the listing in the index and open the .torrent in their client. Client contacts tracker which, in short, returns list of seeds & peers. Client contacts seeds & peers and individually tries to hook up all connections it can. Enter DHT, which means, I think, is that you get peer lists from peers you already are in touch with.

***

Now it turns out that a pretty famous public tracker went down a couple of months ago, but with some lingering hope of its resurrection still present. I found a torrent on one of the *novas, with the only tracker listed being that of downed-site. On a whim, I open it in utorrent, and although tracker status is "No connection could be made because the target machine actively refused it.", I get some seeds and the torrent is "working".

My question: where did it get the list of seeds from?

If it's from the downed-site, is the moral that it's "down" but not really?

posted by daksya to computers & internet (6 answers total)

Your client is probably using the DHT (trackerless) network.
posted by Memo at 7:56 AM on January 12, 2008

Wikipedia indicates that the DHT used in Bittorrent is Kademlia.

In the Kademlia entry, it says

"A node that would like to join the net must first go through a bootstrap process. In this phase, the node needs to know the IP address and port of another node (obtained from the user, or from a stored list) that is already participating in the Kademlia network."

Where did my client get that first IP from?
posted by daksya at 8:10 AM on January 12, 2008

Often, the tracker is on a completely different server than the website where you download the .torrent files. If the site you're referring to is the one that I know of (which I miss dearly every day), then the reason you can still get peer lists is because their tracker wasn't taken down; only their main site.

There was talk in the site forums and other places online about how it wasn't really taken down by the CRIA because of this fact, but that turned out to be not entirely accurate, as they were taken down, just not the tracker server.
posted by omnipotentq at 8:14 AM on January 12, 2008

There is just one DHT network–it isn't per torrent. So if your client knows the IP of at least one other person on DHT that it's gotten from other torrents, it can contact that IP and rejoin the network.
posted by zsazsa at 8:16 AM on January 12, 2008

zsazsa, so once my client is part of the network, it announces on DHT that it would like to start this torrent and hopes an existing member of the swarm receives the message?
posted by daksya at 8:25 AM on January 12, 2008

Sort of. It'll do a search on the network for the file hash that it got from the .torrent file, and hope that there are nodes "close to" that hash. Once it has a node, it can get a list of IPs from the node, and contact them just as if it got the IPs from a tracker. The Use in file sharing networks section of the Kademlia Wikipedia article explains it more.
posted by zsazsa at 9:05 AM on January 12, 2008