Nice. This does some things very similarly to Hyprspace[1]. The core idea is the same: Receive some bytes from a TUN device, shove them into a network socket, and vice versa. Hyprspace uses libp2p to manage the outer connections between VPN nodes instead of plain UDP, which takes care of addressing, hole punching and encryption.

BTW: You can also use the netlink library to configure the routing table without external processes[2]. The /1 trick isn't necessary either, you can just create a route for 0.0.0.0/0 and set its metric lower than the existing default route. That won't replace the old route in the table, the new one will just take precedence as long as it exists.

[1] https://github.com/hyprspace/hyprspace

[2] https://github.com/hyprspace/hyprspace/blob/a5957e485ff0c2e9...

Thanks for the comment. Cool stuff you lot are doing on Hyprspace! I agree the usage of `exec.Command` is a bit gnarly, using the netlink library would make it a bit more cleaner. The suggestion to define a metric lower than that of the existing default route makes sense. But then that would mean I then find out what the default route's metric is by making an extra call where my current implementation does not use that extra call but the outcome is the same? To reduce cognitive complexity, I can understand how your suggestion helps though :), unless there's another thing I am missing?

Ah nvm, I see how it reduces the complexity now. I would not have to create routes for the halved address spaces anymore

Since people are apparently interested in minimal VPNs, here's one I built in Rust recently: https://github.com/atereshkin/nanovpn

My goal there was to have as little code as possible so that one could look at it and immediately grasp what goes into establishing a VPN.

Awesome, thanks for sharing. Did you use anything particular to help direct the implementation? I've been on a streak of building things in Go for the same reason (learning), and a VPN is one of the items on my list.

Hey! Thanks for the comment. There was not any particular one resource I used to direct this implementation, I used a couple. Most prominent was reading open source code on GitHub[1] by people who have done similar things. Understanding how TUN/TAP devices work was crucial and this book[2] helped a lot.

[1]https://github.com/dsnet/udptunnel [2]https://www.amazon.com/Internet-Security-Hands-Approach-Comp...

pretty cool and thanks for the details about TUN devices!

I believe wireguard runs over UDP and while you still need a TUN device, it has kernel implementations to handle encrypting the traffic.

Thanks for the comment and nice words! The encryption bit is still something I am yet to get into since I have not decided on how to go about that yet! It seems Tailscale uses the WireGuard protocol, I could probably look into how they are doing that and get some ideas from there :)

Tailscale actually implements wireguard in go in user space, last time I checked.

netbird is more interesting as it's able to rely on kernel-mode wireguard when peers can see each other directly, and fall back to the userspace implementation.

From the article:

> Currently, packets are not being encrypted within the UDP tunnel so packet sniffing over the internet is possible. It is encouraged to use this over a protocol like SSH

No encryption takes the P out of VPN. Also, if you are going to need SSH to make it secure, then you can just use OpenSSH's built-in support for the tun device using the -w option.

Agreed. I will be working on eliminating that limitation soon. For now, I have removed that part from the blog post to reflect the current situation a bit more clearly.