4 ms·

Considering the goals of IPFS, I’m surprised it has trouble at this scale. I mean, it’s supposed to be “interplanetary” but this data would fit on two hard driv

by wrs 4y ago

Considering the goals of IPFS, I’m surprised it has trouble at this scale. I mean, it’s supposed to be “interplanetary” but this data would fit on two hard drives. The post talks about performance problems with advertising “hundreds” of content IDs. Is it a design problem or just an early implementation problem?

GauntletWizard 4y ago
Design. IPFS is a dead end in design because it's properties are designed with statistical perfection to it's anonymity. This can only ever be achieved statistically, because there's always the attack of "Start Removing nodes at random, and when the content disappears when you remove a node that was your one" This is a provably unsolvable problem, but it's okay: nobody actually has this problem but the worst social outcasts. In some places that's itself a problem: what "The worst social outcasts" is varies from place to place, but running IPFS is enough proof for a death sentence in those places.
- dale_glass 4y ago
  You might be confusing IPFS with something else, like Freenet. As far as I know, there's no such statistical anonymity in it.
  - csande17 4y ago
    Yes, it's fairly easy for attackers to discover the IP address of the people currently hosting a particular piece of content on IPFS: https://docs.ipfs.tech/concepts/privacy-and-encryption/#what-s-public-on-ipfs https://docs.ipfs.tech/concepts/privacy-and-encryption/#what... I think the parent comment is asserting that even if you run IPFS over Tor or something, or even if they try and make IPFS more anonymous in the future, it will always still be vulnerable to the "cut off your Internet and see which node in the network goes down" attack.
ReaLNero 4y ago
Decentralization is expensive e.g. downloading a Google Drive file is immediate while a torrent takes some time initialize connections, or the PoW protocol for cryptocurrencies. Perhaps we could narrow down the purpose of IPFS to storing a small amount of data in a database that is impossible to delete and difficult to censor?

by wrs 4y ago

GauntletWizard 4y ago

Design. IPFS is a dead end in design because it's properties are designed with statistical perfection to it's anonymity. This can only ever be achieved statistically, because there's always the attack of "Start Removing nodes at random, and when the content disappears when you remove a node that was your one" This is a provably unsolvable problem, but it's okay: nobody actually has this problem but the worst social outcasts. In some places that's itself a problem: what "The worst social outcasts" is varies from place to place, but running IPFS is enough proof for a death sentence in those places.

dale_glass 4y ago
You might be confusing IPFS with something else, like Freenet. As far as I know, there's no such statistical anonymity in it.
- csande17 4y ago
  Yes, it's fairly easy for attackers to discover the IP address of the people currently hosting a particular piece of content on IPFS: https://docs.ipfs.tech/concepts/privacy-and-encryption/#what-s-public-on-ipfs https://docs.ipfs.tech/concepts/privacy-and-encryption/#what... I think the parent comment is asserting that even if you run IPFS over Tor or something, or even if they try and make IPFS more anonymous in the future, it will always still be vulnerable to the "cut off your Internet and see which node in the network goes down" attack.

ReaLNero 4y ago

Decentralization is expensive e.g. downloading a Google Drive file is immediate while a torrent takes some time initialize connections, or the PoW protocol for cryptocurrencies. Perhaps we could narrow down the purpose of IPFS to storing a small amount of data in a database that is impossible to delete and difficult to censor?