I'm not really sure. I need a bit more crypto knowledge & math, and a good chunk of time to brainstorm for it; no solution comes to mind. They're all necessarily bound by that you need to trust at some point; but ease-of-use is paramount in my opinion, if you want to actually change things. No matter what, there are gives and takes.

Browser plugins might be the eventual solution's first steps, though they're more and more becoming sandboxed websites (which I like. Fewer security issues, easier programming, etc), so you'd have to go with something lower-level, which means it's harder to do cross-platform. But that's likely to be the case regardless, unless a single platform wins or virtualized, standard OS APIs become the norm.

All that said, I'm not sure there is a best solution, nor one which I'd actually be happy with. Much less something which works efficiently on a global scale. But I'm essentially a communication-anarchist: I generally think it would be best if anyone, anywhere could privately, anonymously communicate with anyone else. And I realize just what a can of worms that would be.

edit: Just for clarification, as I sometimes come off this way: this is meant in no way to be an attack on the idea / goal / you. And if I'm missing something, I'd love to know. Discussions like these often lead to solutions though, so I enjoy them and end up saying a lot :) I think some of it comes from having both of my siblings in debate teams, and having judged at a few debate competitions; I tend to come off more certain / forceful than I intend.

I'm not particularly defensive here, just fishing for people who'd like to help on the idea - which I came up with right on the thread above.

The thing is, the dns-backtracking-browser-plugin sounds like a simpler and more doable approach compared to anything else I've heard of. Any more elaborate approach would have to settle who owns a domain and that's not any easy thing for the present system.

It would certainly need to be system/browser specific but otherwise doesn't sound hard. Indeed, I could do it in a couple weeks and a really smart person could do it in a day.

Obviously it's a stop-gap. The distributed peer-to-peer client featured here a couple of weeks ago is a far more robust solution. (see http://news.ycombinator.com/item?id=1985431). That would include a system fairly similar to what you describe.

I've been meaning to read through the details on that for a while... guess I'll just have to do it.

But yeah, trackerless-torrents are about the epitome of such a system, though I think it'd have to be changed drastically to support a fast query architecture like DNS-like services need.