A convoluted route around t'interweb lead me to AVFS. This looks like the start of "the web filesystem".
This was an idea I had a few years back - it's a filesystem, but structured like the web. That is, you go into a directory called http/www.coofercat.com and you'll find a file called index.html. If you read it, the filesystem makes an HTTP request to that location and gives you the file. If you try to open a file you haven't seen before, then the filesystem tries to fetch the file and give it to you. If you get a 404, it blows you out with a "file not found" error.
The addition of a cache to this filesystem, and you get a layer in the OS that takes away a whole raft of code from every broswer on your system. It also takes out chunks of code from all the libraries and programming environments you have. It means that any old program can move in and out of the internet when ever it needs to. Imagine every crappy old Bourne shell script you ever wrote becoming Internet enabled?
This could give everyone a really easy, really clever remote filesystem. Imagine working at home on your laptop, and saving some stuff to your office server. You don't need a VPN, because the filesystem uses SSL and authentication. Next, you get a coffee and get on the Internet and change that file on your way to work. You arrive at work, and tell your boss, who's abroad, but he helps himself to the file from his hotel room.
Ahh, the possibilities. There are a load of details to get around, but I reckon it could be really powerful.