Hacker News Reader

Comment by sharemywin | Hacker News Reader

Is it me or is the elephant in the room like "get paid to put pirated movies, games, music, software on the network"

Obviously, no one is going to admit it because then your network is tainted but every file sharing outfit ultimately ends up do pirated.

I've been mining lbry and it seems lot more ahead of these others. https://lbry.io/

View on HN

We're building the basic building blocks of autonomous software. If Ethereum is universal compute, Filecoin will provide the storage backing for it.

The beneficiary will be AIs, not humans.

Right now, we're just now thinking of clouds, containers, and "serverless" (functions-as-a-service). If we take those ideas and iterate them forward, we converge closer to a real "serverless" paradigm -- what Ethereum already provides today in principle, if not at scale. This will require storage that can keep up with it, and that is where FileCoin comes in. While human consumers may benefit from this, it enables autonomous software that can live on its own and manage its own matabolism (resources accounted for through the blockchain). What those autonomous software will do? I don't really know.

The https://lbry.io/faq/what-is-lbry is just as empty buzzword bingo as the Filecoin docs.

You can't create a business out of pirated content (anymore). You can maybe create such a network, but not gain money or stay out of prison.

Filecoin to my knowledge has or will have no capabilities of sharing files. It's dumb end-to-end encrypted decentral storage (which is a good thing!).

so you have file storage and a currency. you need a dao to manage the index/search.

dao=data access object?

You can't implement indexing and search over end to end encrypted data file chunks. You would have to work with the hash of complete files and a localizer, and you want that in the initial design.

Distributed org creates account rents massive file store builds index and use contracts to enforce rental fees.

Thats auper nice

For OpenBazaar, built on IPFS, we've gone the route of federated search with an open API. A DAO could be interesting but a lot more work for little gain.

Eventually I'd like to open source code that lets any user create their own index/search server and potentially charge for usage if they want.

Filecoin is being built on top of IPFS (ipfs.io) which, in principle, could be used as a really robust network for sharing pirated content.

However, the IPFS team is aware of this problem, and is building in measures to allow nodes to opt out of storing globally available blacklists of known bad objects (i.e. pirated content or illegal content)

In practice, this is definitely a really hard problem to solve, and very important. But I think doable.

If you're hosting something such as a movie, and it's blacklisted, couldn't you just add a small amount of noise to the video file to make it unblacklistable?

I guess that depends on how things are being blacklisted.

Exactly. YouTube has the same problem and does all kinds of things to detect this. They have the benefit of already having the media on all their own storage which they can run a map-reduce on, however.

For now, but this sounds like a problem machine learning was built to solve. Youtube has issues because with the sheer volume of video they don't have the silicon to transcode/examine every frame, but on a more limited network with less traffic such analysis might be doable. The more "variations" people introduce in response, the larger the training set for the algorithm.

Yes, this is an adversarial machine learning problem. I highly doubt those kind of measures would be implemented by file coin though, considering the cost of scanning files for such at scale. I'm sure, at least in the short term, it will likely be some time of hash or edit distance measure (or an equally naive and easy to fool approach).

What if I just encrypt the video before I put it on the network and use a side-channel to distribute the key? ML gonna have a tough time with that.

Since the content is encrypted, how would you blacklist any content? The nodes don't know what kind of material they are hosting.

You don't blacklist based on content, but on the hashes of the content, which the nodes have to know. At one point or another, the decryption key has to be shared for the encrypted content to be usable for anyone, and if the copyright holders get a hold of that key, they can prove a violation and blacklist the hash of the content.

You do realize though that changing just 1 bit can change the entire hash? What's to prevent people from appending a random bit or two, and getting an entirely new, unblocked, hash?

Responding to specific take down requests is enough to activate the safe harbor provisions in the DMCA. Legally they don't have to do anything else.

I was just explaining the general concept. I don't claim to have in depth knowledge on the subject.

I think blocking a hash can already be a victory, since it basically makes all the metadata that was associated with the hash useless and forces reseeding under a new hash and publicizing that hash. As long as you make it inconvenient enough, very few people will bother, which is the best the copyright holders can hope for.

Reseeding might be pretty easy though in IPFS since there is a block layer below the object layer, and I guess you can't really block block-hashes.

Great point; this happens a lot with YouTube videos, and would definitely get worse with other sorts of media

Your plan is to create a centralized list of things "good and moral people" should avoid, that only good and moral people will opt-into? That's a funny strategy for several reasons.. First - do you really want "bad guys" to have a list of all the content they shouldn't have? Might as well call it do-not-read.txt. Secondly, if the system is opt-in for "good and moral people", will that really do anything to stop the distribution of content on do-not-read.txt among "bad and immoral" group?

I'm personally glad that the fathers of the internet had better things to do than moralizing legality.

Its not about "good and moral" vs "bad and immoral", but "legal" vs "illegal". Blacklists are not currently implemented in IPFS but one imagined a solution, where people opt-in to their e.g. state-supplied blacklist to prevent run-ins with the government, which could say "if you aid criminals by providing them with illegal content, we will hold you accountable for it". In the end, it is probably nothing more than a facade to appease governments that in the end will come up with something similar ineffective.

you mean, illegal in Russia? Or Uganda?

Depends on which risks/rewards you want to take on, just like today. If you live in China, you might want to adhere to the Chinese blacklist, but outside of it, you probably don't care, since the Chinese government also doesn't seem to care _that_ much.

> since the Chinese government also doesn't seem to care _that_ much

Sounds like the opposite of something most would consider "important".

So basically, Pied Piper plus a coin?

And without compression. And precisely what sia and storj do

AFAIK the uploader has to pay in this type of system; it's basically just S3 but cheaper and "decentralized". You can upload pirated content to a public S3 bucket if you want and I don't think AWS really cares because they're getting paid and they are covered by the DMCA safe harbor.

If the society is collectively willing to pay for storage mostly to host pirated content, that tells you a lot about what the real (rather than publicly expressed) social consensus is wrt piracy.

Last time i checked my local usenet, it was alive and thriving

It's not, the real business model here probably looks a lot like rapidshare.

It's more about enabling applications with decentralized data storage than it is a new type of object storage.

Reply on news.ycombinator.com