As a counter argument, your standard browser already caches a ton of stuff, might as well serve some of it back up if your upstream can support it. If you replace your browser cache with IPFS, and replace bookmarking with pinning you've got a functioning system. Your hosting requirements for content will likely go down, you get massive scale redundancy and reliability for free, no more slashdot effect and you might be able to claim 100% uptime due to just how unlikely it would be that someone hits your site while your root host is down and it's so low traffic that it's not bouncing around the network, even if it's just a standard consumer phylink.
You could still get 'a hug of death ' but as the host you can just sit back and let the infrastructure work, distribute and recover, ala what it looks like to be the first seeder of a torrent. I predict this is unlikely in real world situations though, as for someone to share your site they'd likely have visited it first, thus your content host is likely to not be the only host. I'm an optimist, mostly because I think is stupid where we are at concerning self hosting from our homes.
Re:Putting This Blog on IPFS
by MrEldritch on Wednesday September 19, 2018 @08:35PM
(#18028797)
IPFS is basically a cross between BitTorrent, plus Git's datastructure for permanently and uniquely identifying particular versions of and updates to a given piece of content, plus a clever distributed hash table for keeping track of who has what content.
Re:Putting This Blog on IPFS
by bordercases on Wednesday September 19, 2018 @09:45PM
(#18029128)
Am I the only one who finds this prospect very exciting?
Having said that, I think IPFS sounds like a really good idea -- oddly, it's a lot like what I envisioned in a science fiction novel, where the notion of data being replicated across hundreds of storage nodes is so taken for granted that the main character has trouble conceiving of the notion of data that has a "location" (that is, is stored on only one device). But to get there, it needs to be largely content-agnostic: if the data is out there on the network, then it's replicated across those hundreds of storage notes, regardless of popularity. IPFS proves that technology is basically already here in theory -- but in practice, I'm not sure it's feasible in terms of storage costs/requirements yet.