With IPFS every single website you look at becomes cached by your node and redistricted by your node, that’s the whole point of it. Redistribution is illegal by default, unless explicitly allowed or public domain. The problem is even if it is allowed, say Open Source software, that often comes with conditions such as “you must include the license when you redistribute it”. With IPFS even that doesn’t work, as each file or even subsections of a file will get redistributed independently, so if the license is in another file than the one you are redistributing, you are in violation of that license. With Bittorrent in contrast you redistributed whole directories at once, so that’s fine.

Unless you want to use IPFS exclusively with only 90+ year old works with expired copyright, I just don’t see it working. At the moment nobody really cares, since it is small enough, but that can quickly change.

ISPs and sites like Youtube have exceptions that allow them to redistribute illegal stuff, if they remove it when they are notified. No such exception exists for regular users and I’ll doubt that we’ll ever get one, as with IPFS there is no origin of a piece of content that you can shift the blame to.

It means that using it properly is automatically illegal. I am not seeing how that’s a “feature”. It renders it completely unusable.

Proper data model would be a start, i.e. public-key based identities instead of just the old name@server. That way you could hop from server to server and still be the same account. Would make the whole thing a hell of a lot more robust, as in case of server failure could just continue as if nothing happened on another server.