In general, I don't think a blockchain is the right data structure for decentralized file hosting.
Me neither, that's why I think this proposal, filecoin included, is fundamentally flawed. If I want to share a file, I'm the one who first of all have to take personal responsibility of the file, and second, pay the required cost. There's no reason a hundred miners keep my file when I only need one to store it. That's the definition of inefficiency. If I want true decentralization of my file storage, then I should run a BitTorrent client, and seed my stuff there. If I don't care about that (which I honestly don't), I can stick with a centralized solution, such as Mega Upload. Privacy can be assured with encryption.
And that's before I even comprehend (2 pages now) how Proof-of-Work can help the situation in decentralized file sharing / storage. It's just flawed, no need to. Blockchains solve the double-spending problem, and that's all they do.