That's true, but I2P is a weak anonymity network with little research, some of which recently revealed that users can be easily identified by attacks on the netdb: Practical Attacks Against The I2P Network http://wwwcip.informatik.uni-erlangen.de/~spjsschl/i2p.pdf BitMessage has serious security flaws as well: https://bitmessage.org/forum/index.php/topic,1666.0.html Nym servers are ok. kmf has talked about some the PIR methods that are information theoretically secure, so as long as they are implemented in safe ways, they can be a good alternative. I hadn't heard of tox.im so quickly read up on it. It hasn't been released yet, but hopefully soon. It looks interesting in that it uses DHT, but it does not appear to be anonymous, ie it relies on finding a user's IP address. This is bad for most people in this community, who want to remain anonymous to each other. The only way to address people over Tor is with the onion pseudo-domain protocol, so maybe that will be added later. Most hidden services are small and don't take up much memory anyway. Tor networking is CPU-bound because of the crypto operations. Relays max out their CPUs long before their bandwidth. But even if that weren't true, VMs provide orders of magnitude better security, which is a trade off worth making. Some well known hidden services have leaked their IP address because of misconfigured web servers and other services, including our very own Silk Road! DPR is just lucky the wrong people didn't see it. The only way I would ever run a hidden service is in VMs so the service people interact with does not and essentially cannot know its public IP address.