I'm fairly sure this isn't really safe or really reliable.
To my knowledge, IPFS isn't really private, in that both the nodes hosting content can be easily known, and the users requesting content can be monitored. This is bad news for something law enforcement has already taken a serious interest in.
IPFS also requires "pinning", which means that unless other people decide to dedicate a few TB to this out of their own initiative, what we have currently is a single machine providing data through an obscure mechanism. If this machine is taken down, the content goes with it.
The amount of people that have 31 TB worth of spare storage, care about this particular issue, and are willing to get into legal trouble for it (or at least anger their ISP/host) is probably not terribly large. The work could be split up, but then there needs to be some sort of coordination to somehow divide up hosting the archive among a group of volunteers.
> To my knowledge, IPFS isn't really private, in that both the nodes hosting content can be easily known, and the users requesting content can be monitored. This is bad news for something law enforcement has already taken a serious interest in.
You can access the data through a VPN.
If necessary, the hosts can also encrypt the filenames and data so that, until law enforcement gets the encryption key, they can't know who accesses what (public key and other necessary info would be communicated through Signal). Rotate the filenames so so when one is discovered, past requests can't be tracked. Maybe there is a way to slightly break the protocol to further hide the requests.
> IPFS also requires "pinning", which means that unless other people decide to dedicate a few TB to this out of their own initiative, what we have currently is a single machine providing data through an obscure mechanism. If this machine is taken down, the content goes with it.
Do you have to explicitly choose what data to pin? If so then this is an issue. If not, and you just pin random chunks, then if we normalize people using IPFS and distributing legal data this will be solved. If we normalize it enough, there will be too many people hosting and using IPFS for law enforcement to reasonably take down. Or we could just have enough activists that are willing to risk being fined or arrested.
---
That being said though, I'm still not convinced on IPFS because it seems like it cannot handle much and is excessively inefficient (case in point: this article). The authors of IPFS should release a new protocol which addresses issues like the article's, hopefully before too much adoption.
Okay, and then law enforcement asks the VPN. Yeah, it improves matters some, but we're talking about a huge book archive here. People aren't going to maintain OpSec
> If necessary, the hosts can also encrypt the filenames and data so that, until law enforcement gets the encryption key, they can't know who accesses what (public key and other necessary info would be communicated through Signal).
That's a plan suitable for some sort terrorist organization maybe, but exactly how is that going to work for an archive of millions of books that are intended to be served to the general public? What's the key distribution mechanism? How do you distribute keys to everyone but the cops?
> Do you have to explicitly pin the data? If so then this is an issue. If not, and you just pin random chunks, then if we normalize people using IPFS this will be solved. If we normalize it enough, there will be too many people hosting IPFS for law enforcement to reasonably take down.
IPFS isn't Freenet. My understanding is that it's a content-addressable, multi-source system. Meaning the main different thing from plain HTTP is that stuff is named by hash, and that if there's a dozen people serving a given file, then the system can spread the load among them, or tolerate some of them going offline. You ask for hash X, the system figures out where to get it.
Unless people make the intentional choice to mirror content, then it's not very different from serving stuff over HTTP, only with a worse user experience.
What you suggest sounds more like Freenet, but I doubt that it'd work great even there. Freenet does the "store random chunks" sort of thing, but this means that it's extremely inefficient, and easily loses data. Freenet was made for plausible deniability, so any storage is probabilistic, and data is replicated as it moves through the network and eventually lost if nodes go offline or it just falls out of storage due to the lack of interest. Storing 31TB would require a lot of nodes dedicating a lot of storage, and a lot of interest in accessing all of that data on a regular basis.
> If we normalize it enough, there will be too many people hosting IPFS for law enforcement to reasonably take down. Or if we just get enough activists that are willing to risk being fined or arrested.
That's not a great plan for something that already got people into legal trouble
>Okay, and then law enforcement asks the VPN. Yeah, it improves matters some, but we're talking about a huge book archive here. People aren't going to maintain OpSec
You're making it seem like this is such an impossible task, but libgen already uses ipfs as a mirror..
By glossing over important points and focusing on ones you feel you can argue against you’re coming across as someone who simply wants to argue.
I’m only saying this because I suspect that otherwise if people fail to change your mind you’ll walk away thinking you were vindicated by their silence, when it’s just as likely that they’ve followed the old adage: “Never Wrestle with a Pig. You Both Get Dirty and the Pig Likes It”.
To my knowledge, IPFS isn't really private, in that both the nodes hosting content can be easily known, and the users requesting content can be monitored. This is bad news for something law enforcement has already taken a serious interest in.
IPFS also requires "pinning", which means that unless other people decide to dedicate a few TB to this out of their own initiative, what we have currently is a single machine providing data through an obscure mechanism. If this machine is taken down, the content goes with it.
The amount of people that have 31 TB worth of spare storage, care about this particular issue, and are willing to get into legal trouble for it (or at least anger their ISP/host) is probably not terribly large. The work could be split up, but then there needs to be some sort of coordination to somehow divide up hosting the archive among a group of volunteers.