It's nice to have alternatives but I don't understand why we can't have a fully FOSS implementation of overlay networking without needing a company trying to up-sell us to premium features. I don't like to rely on companies to maintain a product that's a critical part of my infrastructure.

I've been using netbird for couple of months now.

The good:

• I was able to do everything using the free version (exit node, DNS)
• in theory you can self-host it (haven't tried)
• it has an Android client app in f-droid
• stable, everything works really nice

The bad:

• it eats up a lot of battery on Android. There's a bug open for it for months now. After playing around with some settings I got it to the point where the battery usage goes up on mobile data only. Once I'm back to WiFi it's ok. I have deployed netmaker locally and testing it is on my to do list but as things are now I can go through the day without issues with netbird and I'm kind of hoping they will fix it before I'll have the time to test netmaker

Netmaker was easy to setup locally but I don't know if it can fully replace netbird.

I tried headscale and it wasn't easy to set up. From what I read tailscale apps treat self-hosted solutions as an afterthought so it's not as easy to set up. I would be worried they will drop it and force you to use their servers at some point. Netbird at least is marketed as a self-hosted solution.

When I was going for a selfhosted mesh tunnel solution this year, I originally steered toward Netbird given that Tailscale (the client) is rather obtuse to connect with alternate control servers.

I don't actively use Netbird, but rather did go with Headscale in the end. If you're satisfied with and your requirements for Netbird don't include anything that deviates from their setup scripts, I'd imagine your deployment would go fine. My requirements and my strong desire to have everything relevant in my selhost deployments be fully to mostly declarative, and be trackable via git had me finding out how incompatible Netbird is for me. I don't think I have seen an app more averse to being deployed declaratively, which as it turns out is likely the reason their deployment scripts exist in the first place.

The there's the (at the time at least) really confusing and contradictory documentation that describe two different generations of potential deployment of the software. Good documentation is half of what makes software deployments maintainable, and Netbird had (and might still have) outright misleading documentation that will have you looping in circles attempting to figure out reasonably standard functionality like using external oidc for accounts.

I've used nebula for a while now, and I'm working on building a little network management UI for myself because I'm so tired of manually creating new host configurations lol

Tailscale clients are open source, controller servers aren't

So the main parts that you need aside from Wireguard support in your OS, is the Tailscale client and the Headscale hosted server?

Is the client fully FOSS? Or are there still any parts of this typical deployment that could be rug-pulled if the company gets enshittified?