Elegant open source project tracking, Trello like but self-hosted

I can give them resource limits the same way (CPU/memory limits, except easier than cgroups), as well as set restart policies and have a clear look at what's executing where, with something like Docker Swarm it becomes like systemd across multiple nodes and scaling up/down becomes easy, especially with load balancing for network calls. Software like Portainer also has pretty nice discoverability.

Speaking of networking, I don't have to worry about tunnels or firewall configuration myself, can just expose a web server that acts as a reverse proxy and give everything else custom private networks that span across nodes (with something like Docker Swarm again, though Consul and Kubernetes have the same functionality, details aside).

I can have custom port mappings (regardless of what the software uses, I might not even care about digging in some configuration file to change it), which is especially useful when running multiple separate instances on the same machine (like different versions of PostgreSQL, or separate instances for different projects), or hostnames in case I don't want to expose ports.

I can easily have custom persistent/transient storage paths or even in memory storage (tmpfs), when I have persistent storage then suddenly backups become easy to do and I can be very clear about all other directories being wiped and being in a known state upon startup/restart. It's also immensely useful for me to escape the sometimes weird ways how software on *nix uses the file system, I can just mount my persistent files in /app/my-app/database/var/lib/postgresql/data or /app/my-app/web-server/etc/apache2/sites-enabled and know that I don't care about anything outside of /app.

I can also treat Docker as lightweight VMs, except a bit more stateless, in that I can have container images that I base on a version of Debian/Ubuntu/Alpine or whatever, ship them, and then don't have to worry about a host OS update breaking something, because only Docker or another runtime like Podman is the actual dependency and most of the other software on the node doesn't come in contact with what I'm running. With rootless containers, that also improves the separation and security there a little bit.

With all of that in place, suddenly I can even move apps and all of their data across nodes as necessary, load balance software across multiple nodes, be able to easily tell people how to run what I have locally and store and later use these images very easily. Are there pieces of software or alternatives (e.g. jails) that do a lot of the same? Sure, but Docker essentially won in ease of use.

That's the biggest problem I see with Docker: nobody has an incentive to make well structured software with a lean dependency chain and a straightforward installation process… These used to be good proxy of the overall software quality of the project, but now Rube Goldberg projects that just happen to work by luck are routinely distributed and the user has no idea of how big of a mess it is internally.

Python has notoriously awful dependency management. One of the biggest appeals of Docker is that it lets you build the equivalent of a "fat jar" so that you get at least somewhat reproducible versions of your dependencies at runtime. For a language with decent dependency management the value proposition is much weaker.

I would ask the same question if you were using docker for a Rust project btw.

IMHO Docker mostly make sense when you have projects that require globally installed dependencies (like C or Python).

I was more looking for use-case as opposed to a barf of all dockers....

"I want to do TASK so here are all the dependencises for you to do TASK and how they will link"

---

And yeah; how do you tink a 3-year old in 2050 is going to be able to setup his dev env? Do you want him to learn binary.