Hacker News new | ask | show | jobs
by heartbeats 1467 days ago
Here's your list - you have to use double newlines:

- torrents work if they are shared more than they are leeched, so it's just easier to have it run in the background. There's very little reason to not use the existing client/server models like transmission or tTorrent

- indexing and search still relies too much on third parties. Integrate magnetico (passive indexing), bep 33 (dht-based scraping) and bep 51 (dht-based indexing) and the user gains an order of magnitude of autonomy because they don't rely on centralized authorities anymore

- is there a possibility yo go further ? Make sharing files easier with bittorrent, or something like that ? Bittorrent clients should help with that

I don't think decentralized indexing is a solvable problem. There's too much crap, and too little in the way of useful signal to tell it apart.

What would be really useful, though, would be a way to separate indexing from the actual torrents.

Let me give you an example: say you're a member of a private tracker, and you see some release that looks nice. You click it in your torrent client. It then tries to find the same torrent on a public tracker (e.g. by hash), and downloads from and seeds to both.

If a system like this existed, then you could use private trackers just for the indexing, and then the computers could be left to do the actual work of finding the data.

Also, as a nice side bonus, this would allow for the "long tail" style of seeding that the more old-fashioned protocols have. If you're not seeding a whole torrent, but just a series of single files, nothing prevents you from just opening up e.g. the 'Downloads' folder to the public, and allow them to download if they can provide the hash of the file. This would make it much easier to find seeds, since the limit usually isn't bandwidth but storage.

2 comments

If you use private torrents, chances are you want the swarms you participate in to be private, unless you're happy to announce your IP to RIAA. I do agree though that the current model in which cross-seeding between private swarms is an afterthought and a pain is not perfect. Clients should handle that automatically without having to resort to hardlinks and separate folders and heuristics to find matching torrents and partial matches and the rest of accidental complexity no one really wants.
There actually are BEPs to manage cross swarms sharing, but I don't know of clients that do handle that. BEPv2 should make this easier though, because each file has their own hashes on top of the top-level hash.
This is only really a problem in the US, and for people who aren't using a VPN.
> I don't think decentralized indexing is a solvable problem. There's too much crap, and too little in the way of useful signal to tell it apart.

I've been using decentralized indexing for more than a year and I haven't needed to go back to ad- and malware-ridden search sites. I find the number of seeders a good enough indicator of quality, but of course that only works for popular content.

EDIT: thanks for formatting, I always forget this double linefeed rule