Hacker News Reader

Comment by aaomidi | Hacker News Reader

What do you think decentralized means? It means that hosting it isn't the responsibility of a single party.

If I put something online, and people think it's interesting and pin it. Even if I go down, the pinned content remains up.

Decentralized doesn't magically solve hosting availability. It just removes responsibility from a single party.

View on HN

I would say there are very few IPFS sites where hosting doesn't still end up being the primary responsibility of a single party. That was the point I was trying to make. If you have to pay someone else to pin it and keep giving them funds then it is still your responsibility. If you have a website that you want to scale and load fast for all your users on IPFS then it will still end up being your responsibility. There may be a few exceptions.

There is no technology possible that removes said responsibility. This is how peer to peer systems work. Your criticism is something that no peer to peer system claims to solve.

Actually, I recommend you check out Freenet [0], a project that's 20 years old. In fact, the Wikipedia page explicitly states:

> Information flow in Freenet is different from networks like eMule or BitTorrent; in Freenet:

> 1. A user wishing to share a file or update a freesite "inserts" the file "to the network" > 2. After "insertion" is finished, the publishing node is free to shut down, because the file is stored in the network. It will remain available for other users whether or not the original publishing node is online. No single node is responsible for the content; instead, it is replicated to many different nodes.

[0] https://en.wikipedia.org/wiki/Freenet

Aren’t collaborative systems like that susceptible to abuse? E.g., an attacker could upload 1TB of noise and then shut down, reducing the network size by much more than 1TB.

Of course, in Freenet, if a file is not requested often it will quickly fall off the network due to user churn.

He said that IPFS hosting is the "primary responsibility of a single party". The whole idea behind peer to peer is to avoid precisely that.

Not really, if you want something to be perpetually available in any p2p network you need to ensure that you're still hosting it. That goes for torrents, ed2k, freenet, and ipfs too.

This isn't true of most blockchain p2p networks, eg, Bitcoin. Anything uploaded to these lasts as long as the entire network does.

The reason this works is it costs some amount to initially print the data to the network, limiting that kind of abuse.

And because there is money involved in making it last as long as the entire network.

Once the monetary reasons drop then so does the data.

It's possible for sure, but with Bitcoin and the overall crypto space steadily growing in value into the trillions over the course of over a decade, it seems unlikely.

I'm not claiming the reality of P2P always matches the ideas behind it.

And surely you're not saying centralized hosting being a strict requirement was what people had in mind for P2P?

I'm saying perpetual availability wasn't what what people had in mind for P2P.

Except for nearly all blockchain-based p2p networks.

Eg: https://cryptograffiti.info

Would you call torrents decentralised?

To an extent, although many times you still rely on a centralized tracker and dedicated seeds. Unless you belong to a private tracker where users are forced to maintain certain ratios to continue using a service, then in most cases a torrent will end up like an IPFS site without any pins.

If IPFS provided the ability to track pins from users of a site like private trackers, and provide rewards or some incentive then I think IPFS would gain much wider adoption.

you're describing an architecture similar to a project I'm working on using orbitdb/ipfslog - it's not quite ready for sharing/using but there's more information here: https://www.reddit.com/r/musichoarder/comments/lrqx7m/record...

Thanks. Napster 3.0 here we come ;)

Most torrent clients also "pin" content a user downloads by default, with configuration options to stop seeding once certain conditions or upload ratios are met.

Does IPFS support something similar? e.g. pin this content until I have a 5:1 upload to download ratio

---

I 100% agree on the ability to track pins. I want to see organizations be able to easily peer/pin content from other orgs they endorse; and for end users to be able to follow and help peer for organizations they support.

I want something like Youtube where I choose who appears as in the related videos on my channel and can be a peer to hosting them; and where my users can help to peer my videos (and I can help peer others).

As an end user, my client (or local server/cache) would prefetch the content I follow both to be a peer and to have it ready for me to access.

This would also help with locality of data; if 4 people in my household watch the same video or channel they would be able to get the content locally. If I'm an organization with a physical location, users connected to my network can fetch the content on a local link.

Some torrents. There are many torrents basically held up by a single peer and those are not in practice decentralized. They could become so if more peers start seeding the torrent, but just because the architecture provides for decentralisation does not make all torrents decentralized.

You're speaking from actual experience, aren't you? I tried out IPFS and came to the same conclusion myself. Content has to be kept alive or it disappears forever, and keeping it alive requires some kind of investment.

I like IPFS, I want it to succeed, but HN has these romantic ideas about IPFS that don't quite match reality, and you get downvoted for being honest sometimes. Not acknowledging the limits of IPFS are part of what holds it back...

Yes. I really enjoy IPFS and fully intend to use it in my next project, but there is still quite a bit to be desired. I think the IPFS cluster project may end up being what I'm looking for, but it currently has a few issues that need to be worked out:

https://github.com/ipfs/ipfs-cluster/issues/1226

Ideally, I could provide instructions for app users to follow my cluster, set the amount of content they want to share, track the size of content they've pinned and for how long, and provide rewards/incentives through the app for doing so.. similar to how private trackers work.

Of course. Just using IPFS doesn't make other people interested in your content. But if you have interesting content it does mean that people will be likely to save it.

You can imagine that if YouTube hosted its videos on IPFS then popular channels would be widely pinned so that even if YouTube shut down and the channel author left most of the videos would still be accessible. And much unlike current HTTP mirrors you would be able to find out who is pinning it without having to check multiple archive sites. (Of course you need to know the hash, but remembering hashes is much lest costly than mirroring whole sites yourself so I think we would see lots of these)

I think that in most cases it is simplest to think of IPFS as torrents. I think the biggest improvement that IPFS made was a single namespace for content instead of each torrent basically being its own swarm. This means that you can find content no matter how it is packaged and you can share blocks between "torrents".

Reply on news.ycombinator.com