Hate that you’re downvoted for this. Tailscale is incredible software but it is buggy as hell.

I gave up on 90% of the fancy features and just do most my routing from one node with good ol’ nftables and ip route/ip -6 route.

I’ve basically got everything thrown into a Samba share on Linux then most media is consumed via the Infuse app for iOS, macOS.

As for music, I have some lossless/hifi that I can stream via Apple’s Music app too.

This is one of those things that if you really want to do it, you’ll have to live with the consequences.

I’m an American that VPNs everything first to my VPS then down a double hop commercial VPN tunnel that finally exits in Switzerland. DNS traffic also travels over that VPN tunnel so you’ll rightly guess that my DNS is rather slow too.

What I do is I run a resolver on the VPS (physically near me) that aggressively prefetches commonly queried DNS records. After years of using Unbound I found Blocky to be much, much faster (especially with huge blocklists). It’s to the point now where sure, it’s slower than a “normal” internet connection but it doesn’t feel slow to me anymore.

Storj Tardigrade with client-side encryption. I use rclone so you could even encrypt it before hitting the Storj library if you’re extra paranoid (among other things like caching, chunking, etc).

It can be specific to the web application but generally speaking you’d want to pass the protocol, client IP address, etc.

Might not be exactly what you’re looking for but I find Gotenberg quite handy for related activities.

I self-host web apps I write myself? ¯\(ツ)/¯

It’s mostly about performance. Caddy’s Go-based garbage collector starts to negatively impact performance at high load. It looks something like:

server {
    listen 443 ssl http2;
    server_name example.com;

    ssl_certificate     /etc/nginx/ssl/fullchain.pem;
    ssl_certificate_key /etc/nginx/ssl/privkey.pem;

    location / {
        proxy_pass http://localhost:3000/;

        proxy_set_header Host $host;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header X-Forwarded-Proto $scheme;
    }
}

I prefer nginx to Caddy myself for reverse proxies. As far as VPN technologies go, Tailscale and WireGuard are where it’s at.

Not sure why we’re comparing Caddy and Tailscale though.

I’m just using WireGuard on a VPS with multiple interfaces. I’m still doing heavy ad/tracking blocking via DNS too.

As for App Connectors I’m working on a script (compiled program hopefully down the road) that can query a specific hostname using a specific interface (say, a US-only website using DNS over a US-based VPN) then create a virtual IP address that directs to that same IP using the correct tunnel.

My reasoning for the virtual IP address is that I don’t want to redirect every website on the host to the other tunnel—lots of servers have an array of websites on them.

What I found disappointing about Tailscale is I had to do a lot of “hacks” to make things work—DNS on each exit node had to match perfectly (despite using different exit tunnels)—then the shit would only work like 20% of the time. One day traffic for the US tunnel worked, the next day it was going out of the exit node. I also never got it working correctly in Docker so I was running multiple VPS servers.

If I remember correctly with App Connectors your client would query the App Connector for the domain, then it would return an IP address. The IP address would be set up to always go through the defined exit node. So if your DNS was off or you were accessing another website on the same server you were screwed. On top of that, it just didn’t work.

I loved Tailscale for about a year but am moving away from it because having multiple exit nodes with each redirecting traffic via commercial VPNs with DNS-based ad blocking and App Connectors grew way too complex.

I’m not saying you’re doing all this but if you do get to a point where you’re directing traffic to multiple countries Tailscale turns into nightmare to manage.

There’s https://mujico.org/

Not being a Python developer myself I’d almost go the Docker route simply to avoid the hell that is Python package management.

While I can’t suggest anything specifically helpful (I’ve forgotten) I’d say check the project’s Dockerfile. It’ll give you an idea of how they’re handling it in Docker therefore a provide some hint as to what to do.

What in the hashtags is going on here?

I resolved the issue and edited the post.

I resolved the issue and edited the post.

I found the domain names (edited the post) and essentially yes, it does route through their servers.

I’ve wanted to for a long time but I want photos to be saved losslessly. I don’t want everything converted to JPEG.

Another vote for Postgres, MySQL kind of blows.

It sure does, but I don’t log my family and friends’ queries so I’ll probably MITM myself using a travel router.