2 ms·

> multiple people have colluded Yes, everyone using OFFSystem is colluding to make it hard to tell who did what. There are legitimate reasons one someone migh

> multiple people have colluded Yes, everyone using OFFSystem is colluding to make it hard to tell who did what. There are legitimate reasons one someone might want to engage in such a system--including privacy. It's the same thing with Tor; everyone with a Tor node is colluding to make it hard to tell who's talking with what other nodes. > it might strain credulity that these "random" blocks "just happen" to have these particular properties The OFFSystem is designed such that blocks do "just happen" to have those particular properties. Any new thing put on the OFFSystem will make use of many random blocks that are already on the OFFSystem. For simplicity, in the examples I gave, storing 1 new data block used just 1 existing data block, but that number is configurable; the default is 2. So when a user stores F in to the OFFSystem, the software will randomly select 2 existing data blocks on the network (S, R), and combine F with them in order to store it: S⊕R⊕F⇒data-block-to-store; the software having having no idea what S and R have previously been used for. For simplicity, I've also been assuming that a file you want to store fits in 1 block; a source block is 128KiB; storing a file larger than that will take storing many blocks. So in all, by default, storing a file will make use of 2×(filesize/128KiB) randomly-selected data blocks already on the network. The idea is that after a short time, virtually every data block on the network will have many unrelated uses. It would be entirely plausible that someone has both a legitimate use for A (e.g. F1) and a legitimate use for B (e.g. F2) with no one involved in that legitimate use having any idea that together they have the illegitimate use A⊕B⇒C, because the whole system was designed to make that likely.

LukeShu 6y ago
Switching sides, kinda: OK, so you might not be able to assign much culpability to any given data block, so just assign culpability to sharing the metadata that A⊕B⇒C, right? It's my understanding that that metadata is stored in a "descriptor block" stored in to the network like any other data block (but isn't combined with a set of random other blocks already on the network). I am unsure what the format of that data is, it might very well be easy to identify a block as being a descriptor block (or maybe not, I have no idea). Going after those who download the descriptor block is problematic for the same reasons--that block will end up being reused for other files, the same as any other data block; they could be using it for any number of purposes. OK, so you go after those that say "hey, use this X descriptor block to get such-and-such copyrighted file". Well, there's a reasonable amount of precedent that that sharing a link to a copyrighted work doesn't constitute infringement in many jurisdictions (e.g. hosting a torrent file or a magnet link isn't infringing; it's actually seeding or leaching the contents of the torrent that's infringing). OK, so you go after those that downloaded the descriptor block, then shortly afterward downloaded the data blocks referred to by the descriptor block, and that's reasonable suspicion. Well (1) tracking that is a lot harder than tracking who downloaded an individual block, and (2) there's still a sliver of plausible deniability, so then then you get a subpoena to search their computer for C. Again, the whole point is to make it really _hard_ to tell who's doing what.

LukeShu 6y ago

Switching sides, kinda: OK, so you might not be able to assign much culpability to any given data block, so just assign culpability to sharing the metadata that A⊕B⇒C, right? It's my understanding that that metadata is stored in a "descriptor block" stored in to the network like any other data block (but isn't combined with a set of random other blocks already on the network). I am unsure what the format of that data is, it might very well be easy to identify a block as being a descriptor block (or maybe not, I have no idea). Going after those who download the descriptor block is problematic for the same reasons--that block will end up being reused for other files, the same as any other data block; they could be using it for any number of purposes. OK, so you go after those that say "hey, use this X descriptor block to get such-and-such copyrighted file". Well, there's a reasonable amount of precedent that that sharing a link to a copyrighted work doesn't constitute infringement in many jurisdictions (e.g. hosting a torrent file or a magnet link isn't infringing; it's actually seeding or leaching the contents of the torrent that's infringing). OK, so you go after those that downloaded the descriptor block, then shortly afterward downloaded the data blocks referred to by the descriptor block, and that's reasonable suspicion. Well (1) tracking that is a lot harder than tracking who downloaded an individual block, and (2) there's still a sliver of plausible deniability, so then then you get a subpoena to search their computer for C. Again, the whole point is to make it really _hard_ to tell who's doing what.