I wonder if Tailscale has ever really been a performance bottleneck for homelabbers

I’m switching off cloudflare for my domain/cdn services to Bunny soon :)

Tailscale just works too well for me and I don’t detect too many enshittification markers yet. But I check in on netbird like monthly now lol

It really stinks that nowadays whenever I find a cool project that achieves exactly what I’m looking for or a solves a problem I didn’t know I had, I have to temper my excitement and look for markers of heavy AI usage (commits, reading md files, the writing and graphical style). it’s really taking the wind out of the hobby, I think. Hopefully it’s mostly just a phase

Those last two bullets would be huge. I have a personal tailnet and another for my org. Switching between them is just annoying enough that I might even pay for that feature.

I always get so close to just setting up wireguard and being done with it. I barely ever change the devices on my tailnet, anyway.

I do have a couple friends on my tailnet to give access to some stuff, so that might be annoying to migrate. That and Tailscale handling all the other networking stuff I might not even know about like cgnat.

ntfy never really had good push to iOS, in my experience. The only way I could keep my private channels consistently working was to use the PWA and specifically not sign into it (otherwise, my login token would expire and break things).

I gave up and switched to pushover and as long as I’m somewhat cognizant about what i’m including in the notifications, I’ve been pretty happy.

I’d love for something self hostable to get as good as pushover on iOS

There’s almost no chance of getting banned for this. Spotify themselves provides the APIs that let you get metadata about your playlists and tracks, and setting up a developer account with them is petty easy.

It’s moreso a concern of Spotify eventually limiting/ratelimiting that data retrieval in the future… which is why you should back up now while you can!

Check out https://github.com/WilliamNT/tunesynctool as well. Its development is a little slow right now but it seems very thoughtfully designed and lets you sync via command line (i’ve done so myself) or you can build around it in python.

Looks cool!

I’m curious about the ISRC matching. I’m working on bringing support for retrieval by ISRC in opensubsonic clients (and Navidrome tends to support the opensubsonic spec) but I didn’t think anyone actually added support yet since it was somewhat recently added to the spec.

I thought maybe it was a Navidrome specific feature to retrieve by ISRC, which would be cool!

But looking at what I think is the track matching algorithm for ISRC seems to just always return unmatched https://github.com/betsha1830/navispot/blob/main/lib/matching/isrc-matcher.ts

Am I just reading it wrong?

Your point 3 would be correct with docker and rootful podman, not rootless podman. I have a whole reddit post where this was hashed out, and over the many months and several comments in the post, I’m fairly certain I’m correct in my stated observations

https://www.reddit.com/r/selfhosted/comments/1ky2j5p/psa_for_rootless_podman_users_running_linuxserver/

That being said, it’s still best to not use id 0 in your containers and mark permissions the correct way based on your system’s user namespace mapping. It’s just one more variable to figure out, where in most people’s case it won’t matter too much, and still provides better isolation than docker

Looks great! It’s tough for me to choose between this, amperfy, and narjo. I think arpeggi has the cleanest UI, but amperfy and narjo have gapless playback, which I really care about. Plus amperfy is fully open source, which is cool. narjo has a bunch of neat features/customization, though, so I’ve been sticking with it lately

https://perfectmediaserver.com/02-tech-stack/os/ this guide uses proxmox specifically but the base distro can be swapped out and the rest of the teck stack and concepts can apply everywhere. I followed it with a ucore (part of Universal Blue) base and it works great.

funny recommendation in a selfhosted community

Maybe Louis Rossman will see this and beat FUTO back into shape? Maybe? Please? I just want one good outcome here!!

I have some guilty pleasure apps that I refuse to get rid of because they’re just so dang good. Many of them are on Apple devices… forScore is one of those apps

Good luck 🫡 I made the switch about half a year ago and went all in on rootless quadlets while I was at it. It was a pretty nightmarish couple weeks figuring out things like user id mappings and rootless permissions, but I got there eventually. Landed on a super neat Traefik config that should work for anyone and makes spinning up new quadlets with their own reverse proxied subdomains really simple. I should really post it somewhere…

In the end I wouldn’t exactly say it was worth it… but it sure feels cool to be fully moved into a more open/native container implementation.

https://perfectmediaserver.com/ This is the guide that really gave me the confidence to take ownership of my self hosting. My setup now is simple (I use ucore from the universal blue project as a rock solid base) and I use small tools that I can understand (snapraid, mergerfs, borgmatic) to make things robust as well as podman and quadlets (docker also works great) to enable easily hosting basically anything I want through containers.

Everyone gave good alternatives for where to buy music. As for the pipeline for getting them into your music server, I have Picard running in a container (a couple projects do this, search docker picard) and I have the settings all configured so that when I drop in files to my NAS (though samba or whatever), then I just double click the folder in Picard and hit save and it moves it into my music server’s directories, all properly and nicely tagged (I have the container volumes all set up properly as well)

You can look into beets or wrtag for more automation friendly tagging services.

Well I can agree on the fact that the arms race situation we’re in sucks. It’s an old problem, seen in malware attacks and defenses. I’m just glad we have people fighting on our side in their spare time :’)

And it’s all good on the tone, thank you for your clarifications

That solution still introduces lots of friction. At the volume and rate that these bots want to be traversing the internet, they probably don’t want to be fully graphically rendering pages and spawning extra browser processes then doing text recognition to then pass on to the LLM training sets. Maybe I’m wrong there, but I don’t think it’s that simple and actually just shifts solving the math challenge horizontally (i.e., in both cases, the scraper or the network the scraper is running on still has to solve the challenge)