A fully decentralized torrent search already exists (and it's in fact funded by public institutions!): https://www.tribler.org/

What do you need a blockchain for? Storing millions of no more seeded torrents?

There are already magnet links and DHT which is distributed shared database. The only issue is a search function, which is much easier to implement and works much faster when it's done in a centralized form.

For distributing all those magnet links of course. Yes, it will be a huge morgue of dead magnet links but not being able to remove them is a required feature. And more importantly it needs to store ratings and not be gameable, or it will be taken over for malware and spam and die a quick death. Which is going to take some thought.

I still don't see how blockchain helps with distribution and what incentive would be to store huge loads of this type of data locally.

Ratings are important indeed, but blockchain doesn't make them not gameable. For this you need some working web of trust and oh how I wish we had one already (yes yes I know the technology is there and it 'works', but it must work like an e-mail works for an average user to really be popular, I want to be warned not to buy from given ebay seller because friend of my friend of my friend had a bad experience and because I assign high weight to this friend opinion, well that went OT a bit..). On torrent sites there's a thin web of trust of website owners moderators and then trusted uploaders, we could do so much better.

Not to make a separate post, small ask HN: any interesting active projects trying to build such user friendly distributed WoT?

In the context of something like the warez scene the ratings system can be read only to end users.

When releases are made to the decentralized torrent repository they are signed by the release group.

Whether or not a particular signer is an official member of The Scene is decided by consensus of The Scene. The web of trust is very small. Everyone involved is tech savvy, motivated, and already knows each other.

From time to time updates to the list of known groups and their keys may be posted, authenticated by a multisignature scheme.

Then clients can automatically discover who's trustworthy and filter on that. No technical skill or effort required of the general public.

You could host the UI in ethereum dapps for example, or IPFS for that matter

Torrents are ephemeral. The persistence provided by a blockchain is totally overkill and comes at a very high cost in resources. Blockchains also don't inherently provide strong sender anonymity. That would have to be built on top.

I think Riffle may be the most sensible option I've seen for sharing magnet links or torrent files. In theory, Riffle doesn't require a large number of servers and doesn't demand unreasonable client bandwidth to deliver protection against a very powerful adversary.

It's just an academic system with a basic proof of concept implementation. Quite a bit of work needs to happen to turn it into a usable system.

For the data itself Riffle is too expensive. Something with a weaker threat model that costs less resources is needed, such as Aqua or Herd. Those systems are also academic at this point.

There are 3 other issues: hosting the UI, DNS, and monetization.

You can get away with hosting a client-heavy JavaScript interface on ipfs as well, but the URL will be ugly making you reliant on Google not dropping you (central point of failure).

And then there is incentive - even if you manage to anonymously take Bitcoin for ads or what have you, someone will eventually take all the work and strip them. The whole thing will have to be a labor of love. Any lovers out there?

> but the URL will be ugly making you reliant on Google not dropping you (central point of failure)

How many people use google to start a search for torrents?

I stopped doing so because the higher results would normally be clickbait links

The UI - couldn't it just be a general client, much like uTorrent or Magnet is?

Yes that's true, and given that torrenting is generally reliant on installed software as well, it shouldn't heavily impact adoption.

I did think of another issue: search, sort, voting, etc. Ethereum is too expensive, too slow, and possibly not capable of that kind of large data set computation - not aware of any capable alternative.