mieru v2 and beyond

[enfein]

Recently there have been reports (issue #50, issue #53, Revolt Chat) of network block with both TCP and UDP protocols in China. It means the current mieru protocols are insufficient to provide a stable proxy services. Even though the original design of mieru has taken a plenty of lessons from shadowsocks and v2ray (vmess), new techniques have been deployed by GFW to identify proxy services. A few known weaknesses of the current protocols are listed below:

1. A full encryption protocol has higher entropy than normal packets. This has been discussed on Sharing a modified Shadowsocks as well as our thoughts on the cat-and-mouse game net4people/bbs#136.
2. The RTT among the first a few packets have special timing characteristic. This is due to the fact that socks5 handshake is faster than dial and connect to the destination server.
3. Some popular websites has deterministic number of connections and timestamps. This information needs to be hidden by multiplexing.

My current plan is to resolve 1 quickly in mieru v2 protocol. v2 means that it breaks backward compatibilities and you must use the same major version of the server and the client. To address 2 and 3, I have an ambitious plan that will take a long time to develop, so it will probably become mieru v3 at a certain point of time in the future.

Thank you for the support.

[mrmmg]

Go on, 💪