All of its install steps other than docker compose are still containers, or relate to hosting services, or specific managed platforms.

I’m a simple admin who likes to compile, copy to /usr/bin, prep my system user, populate a directory in /etc with all necessary configs, and write my own .service files.

If you like running single binaries in their own Docker instances, more power to you, but it rubs me the wrong way when people try to force that on me.

NB: I’m not above reading a Dockerfile and executing the steps manually, case in point: that’s how I build lemmy_server for my current instance… but the Immich steps are waay too involved, with too much of the software expecting some less than sane defaults for me. I went through it once, left a bad taste in my mouth, so I stopped using Immich.

Edit: Just to be clear, here’s the installation steps for Photofield. That’s what I like to see. “Download a binary, run, config lives at xxx” - delicious. I also don’t mind doing a ./configure and make if they want me to - just don’t bloat my system with an entire duplicate OS.

I run Debian on my VPSes. But, also, I’ve been through the path of doing what the PKGBUILD does, which is read the Dockerfike and manually execute the steps: I think it’s pretty problematic to have to do that every update cycle. Got Immich set up, looked at what I’d have to do every time I want to run an update, and "nope"d out.

With server side software like this, which will usually be running behind a reverse proxy and will connect to a postgres dB (or other services like Redis), I kind of don’t want to fight the software/make guesses to create these connections. I also take umbrage with the nginx proxying “example” configurations - just tell me what needs to be proxied to where and let me handle the rest; it’s a server, you don’t know my setup. Case in point, I don’t use nginx.

I like photofield because it’s a single binary (thanks, Go) listening on a port. I proxy what I need to it, and let it run.

Final note: I’m not knocking Immich, or the people who run it. I just think it’s lazy to distribute software with Docker because it encourages devs to write what I consider to be poorly designed software (E.g. Software with hard coded paths as strings, because everything is expected to be in the same place on every system - basically very rigid defaults). As a guy who knows his systems well, it annoys me I can’t manage them my way and need to run a nebulous docker container set up by some web developer. (some shade thrown there, my bad)

You don’t have to agree with me, I’m Don Quixote and this is one of my windmills.