I keep thinking about the beauty of the BitTorrent protocol and wonder why there is no longer any excitement about it and no one is building anything on top of it. Isn't it already a inter planetary file system?
- Residential Internet connections are asymmetrical. That's a problem and renders many clients as leeches.
- It is unreliable for long tail files. If a file is never read, it will never be replicated and will eventually disappear.
- Once the universities started cracking down on BitTorrent usage, it was game over. Dorms were practically the CDNs of the torrent world.
It's also way more efficient to use CDNs. I doubt the internet has the bandwidth for a decentralized Youtube.
Imagine (granted, a ideal world and maybe not actually possible) a world where every routes and computer is a node that both serves and receives data. Suddenly, your ISP can start to aggressively serve traffic from their edge-nodes. If a video goes viral, your local network can fetch it directly from your neighbors network instead of reaching out to the internet to fetch it. We'll be reducing the traffic massively.
But yes, it's a ideal world and probably not possible to execute in reality as the market forces behind paying for bandwidth is so strong.
My understanding is that Hypercore is the successor to dat and also operates along similar lines, but I'll admit that the details are over my head. https://hypercore-protocol.org/
Beaker Browser originally built on top of dat, underwent a significant rewrite and now uses Hypercore instead. https://beakerbrowser.com/
If it works how I think it works, then I can pay a one time fee to host my application, migrate the few APIs that my next.js app is using to be purely onchain Solana programs, then figure out some DNS solution to point it to the right place, and then I don't have to pay monthly hosting, and users can be sure that the value my app provides won't be taken down because I decided to no longer pay for hosting.
It's all theory at least for me right now, but I think it is possible.
I'm not talking about filling it up with useless or silly content though, running a Chia node will cause a hard drive to fail after a very short time. This isn't due to any legitimate usage for storing content, it results from how they implement proof of space. It's been a while since I read about this in-depth, but basically it amounts to having the nodes fill the drive with junk data and then prove they're participating by regurgitating some section of that data on request. Then on successive blocks, it gets rewritten with new data.
The irony is, I believe this was intended to address the environmental concerns of proof of work mining, and perhaps it does. But turning hard drives which require resources to create, into landfill, isn't really environmentally friendly either.
edit: To clarify, I'm not judging people using or participating in the Chia network (many participants apparently didn't even know it would destroy their hard drives). I am judging Chia's consensus mechanism for being wasteful however.
"Plotting" is the act of filling the hard drive with the random data. It only needs to be done once, after which the data can be used to "farm" indefinitely.
Farming is very low on energy requirements and doesn't damage the drives.
Plotting can be accelerated by doing it on a fast SSD, and transferring the plot data on large capacity HDDs for farming. This saves time at the cost of writing a lot of data on the dedicated plotting SSD, which trashes consumer grades SSDs if done continuously.
Plotting can also be done on the HDDs themselves. It's slower but won't noticeably reduce the HDD's lifespan.
When Chia was launched, there was a lot of speculation and farmers were competing to be first to finish as many plots as possible, so most were plotting on SSDs, and many on consumer SSDs. That's were the "Chia destroys hard drives" myth came from.