[syrrim]

>2.

I have thought fairly hard about compatibility. What you could do would be to serve a standard website over http, but allow anyone to act as a cdn. The only centralized piece would be dns to route people to their closest peer. Most clients would think the peer was a normal http server, under control of the website owner. In reality it would be any random person on the internet who had decided to rehost the content. Therein lies the problem: how do clients ensure the integrity of resources if they don't even know to look for it. One of your cdns could be serving fraudulent content, and none of the clients would question it.

In order to distribute http, you would therefore be forced to get at least some custom code into browsers. At that point, why not shove a whole ipfs implementation in?