Some things I've been thinking about regarding the right i2p client model. I'm awful about posting fifty questions at once, so if anybody thinks they have an answer to even part of one of them, feel free to pick and choose.
The usual Tor client model (TBB<Tails<Whonix<Qubes<Physical Isolation) has to change somewhat with i2p.
1. How secure/insecure should we all view the i2p Java router process as? It's definitely miles more code than Tor, and it's Java. Should it be isolated from the host machine more than the Tor process is?
2. i2p is easy to mix with Tor. A proxy server (privoxy/etc) can send *.i2p requests to the i2p router, and everything else to Tor. Or a browser can do it with FoxyProxy or the like. Are there anonymity implications to being on both networks at once?
3. At the point where users are staying connected to i2p 24x7, what's the right setup for clients? How much should they isolate?
I almost wonder if, given the complexity of dual i2p/Tor for many users, it's not easier to pick a cheap hardware platform (Raspberry Pi / CubieBoard / BeagleBone Black) and build a middlebox Tor/i2p node for physical isolation.
Has anybody played with i2p bote? I don't care at all for the idea of the messaging app on the i2p router, but it looks like a decent Kademlia DHT-based messaging system. Just looks like it'd make isolating the i2p router complicated.