I didn't say it was solved, but they are very close. IPFS is essentially a global p2p git repo, with a cryptographically controlled branch namespace.

That sounds very cool. I was just pointing out how divorced these ideas still are from the way we actually program and implement things on the web.

In an ideal world, a link to a source code repository (or a link to anything for that matter...) would never fail because there would be automatic mirrors to at least provide read-only access to it.

It sort of forces one to ask whether this is the result of fundamental mistakes within HTTP / DNS itself. It's not realistic for a web designer to put time into making external links fault tolerant by running some query to switch the routing link to an available node.

Content addressed links at least make this possible. With http you have to reach a particular server. If that server is down, the link is broken. With ipfs all you need is at least one machine on the entire network to be serving a file and the link will work and as an added bonus you can verify the hash of what you get so a mitm attack on that link is impossible.

Very interesting. Do you think ipfs could be used to mimic a server to host a static website that could be loaded by a browser?

You mean like this?

https://ipfs.io/ipfs/QmNhFJjGcMPqpuYfxL62VVB9528NXqDNMFXiqN5...

:-) IPFS includes an optional http gateway to give the traditional web access to IPFS before browsers support the IPFS protocol natively.