I've been working with IPFS a lot more recently thanks to their efforts to snapshot Wikipedia to get around censorship in Turkey. This immediately made me realize the advantages of learning the system, and once you wrap your head around content-based addressing, everything makes sense after that.
I think the biggest thing they're going to struggle with for adoption is their addressing system, which is brilliant from a technology point of view, but impossible to memorize for regular humans. Some sort of naming system (or, ideally, several such systems) needs to emerge I think, before the network will be properly usable by casual folks.
No longer than the corresponding HTTP version: http://jes.xxx/
Of course, DNS isn't distributed or trustless, but if it can work with DNS it can work with the likes of Namecoin or ENS (https://ens.domains/).
Currently you need to prepend an IPFS gateway URL to the IPFS path in order for it to be backwards-comaptible with existing browsers (e.g. https://ipfs.io/ipns/jes.xxx/), but I don't think it will take too many years to get native IPFS support in browsers.
I'm really really excited by what's going possible already, and what's going to become possible soon, with IPFS.
https://ipfs.io/blog/24-uncensorable-wikipedia/
I think the biggest thing they're going to struggle with for adoption is their addressing system, which is brilliant from a technology point of view, but impossible to memorize for regular humans. Some sort of naming system (or, ideally, several such systems) needs to emerge I think, before the network will be properly usable by casual folks.