Ipfs copyright

https://www.reddit.com/r/ipfs/comments/w8gjhw/is_ipfs_doomed_without_tracking_authorcopyright/ (https://archive.ph/wip/ANDLr) copyright HTTP has structure, meaning you can generally go up a level and find further information, on ownership, copyright, licensing, where to send your DMCA requests and such. Some countries have this even in law, e.g. in Germany every website must contain an impressum. IPFS can't do that, individual files can be pinned by itself without the overarching directory structure. Even if you are really careful and only pin top-level directories with copyright information, anybody can still link directly to a CID down below that you are now serving without a back link to the top-level directory. That alone is enough to violate a lot of Open Source licenses. -- I could be wrong--I haven't thought deeply about it--but I don't believe HTTP itself has the structure you're thinking. Websites might have that structure, where you can knock off part of the URL or part of the domain name and request pages up a level, but http only is the communication protocol that may deliver webpages. It is not the webpage itself. IPFS would be similar. It's only the communication protocol, although through its native datatype support it can include MORE structure than HTTP

HTTP points to a host and a host points to a DNS entry and a DNS entry has a contact address in the WHOIS database. Furthermore, some countries have legal requirements beyond that, e.g. in Germany there is "Impressumspflicht", meaning every website must have an easily reachable legal address, random examples: Meanwhile websites that host third party content, Facebook, Youtube and Co. will always have a DMCA link with contact information, so people can send in DMCA takedowns. Basically, with HTTP all content has a clear owner that is responsible for it. Companies like Google can shield users from being directly responsible, but only if they follow the DMCA and remove offending content on short notice. With IPFS that falls apart. Content on IPFS isn't owned by anybody, there is no difference between my uploads and your pins of other peoples content. As far as IPFS is concerned, they are all the same. Furthermore IPFS does not enforce structure, if you build a website, following all the laws, I can just link to a subdirectory of it and your legally required "Impressum" will become invisible. In IPFS I can address individual files or even parts of individual files and there is no parent directory to go up to, there are only child directories. In HTTP I can't do that, a link always follows a hierarchy, I can't link a file without specifying the host. - Exactly. So HTTP doesn't do all that stuff, doesn't enforce that structure. Other protocols and systems may provide it, just as they can if IPFS is being used instead of HTTP. No, with HTTP the content doesn't have a clear owner at all, just as you pointed out, it takes DNS and WHOIS and legal filings and whoknowswhatelse to even give a chance at finding a clear owner. And all of that is not guaranteed! Now, there are an awful lot of boneheaded laws out there, plenty that don't even clearly work with the internet as it is, as the legislators passing the laws don't seem to understand what they're writing laws about. It may very well be that IPFS is illegal in different jurisdictions... but then, I'd say HTTP itself might also be illegal, as bad laws run into practical realities.
 * https://blog.fefe.de/impressum.html
 * https://www.heise.de/impressum.html

IPFS is a bit different from HTTP, but not in the ways you're focusing on. At the end of the day, every bit of content hosted through IPFS resides on a computer that's owned by someone, that's serving it up.If your node is serving illegal content, it's serving illegal content. Your ownership of your computer is pretty clear, and IPFS doesn't do anything to hide that it's your computer serving it. In fact, it may make it even easier to track down the people who host illegal content.

HTTP doesn't make the users the content provider. There is a very clear distinction between the people that consume content and that provide it. IPFS blurs that line. So you argument essentially is that we should never pin or bitswap anything on IPFS to be on the safe side. Great. Why again should we even bother with IPFS when we use it just like HTTP? Yeah, just like BitTorrent and look how well that worked out. No browser even supports that protocol because it's just to dangerous to let lose on regular users, who are not going to ask a lawyer before downloading a file.
 * I'd say HTTP itself might also be illegal
 * At the end of the day, every bit of content hosted through IPFS resides on a computer that's owned by someone, that's serving it up.
 * In fact, it may make it even easier to track down the people who host illegal content.

I wouldn't say IPFS blurs the line between consumers and providers. Instead it flat out says that anyone running a node is both a consumer and a provider, AND is even broadcasting that fact to the world. If a person is consuming illegal content over http, that's a problem, right? At least legally. Well, no different with IPFS, except that because of how IPFS works, it will make it even harder to hide the misbehavior. (At least, assuming the propriety of the laws here.) You seem concerned about copyright type issues, right? Well, with http the authorities would need access to servers, server logs, packet sniffing, etc, to police the downloading of pirated content. With IPFS the IPFS node itself will publicly state that it now has the targeted content, and authorities can address the person running the node. In this way IPFS makes it MORE clear who's participating in piracy. Peoples' computers identify themselves, allowing authorities to automate their reactions, should they want. BitTorrent and IPFS have vastly different use cases and development histories. For one, IPFS is not tailored to bulk transfers such as transmitting a large movie. Instead, IPFS is much more suited to providing a very small bit of data, exactly the sort of data web browsers handle constantly. It seems to me that makes IPFS far more suited to web browser integration since they work together so much better than would BitTorrent.

links
ipfs