What sounds like gatekeeping is often a strongly worded emphasis on having the prerequisite knowledge to not just host your services, but do it in a way that is secure, resilient, and responsible. If you don’t know how to set up a network, set up a resilient storage, manage your backups, set up HTTPS and other encryption solutions, manage user authentication and privileges, and expose your services securely, you should not be self-hosting. You should be learning how to self-host responsibly. That applies to everything from Debian to Synology.

Friends don’t let friends expose their networks like Nintendo advises.

I was pointing out a part of the documentation that you may have overlooked. In good faith, I might add, because the documentation is comprehensive but horribly signposted. Everything I wrote in my first comment is there, but in a less human, more technical phrasing.

https://docs.python.org/3/tutorial/modules.html#packages

The __init__.py files are required to make Python treat directories containing the file as packages (unless using a namespace package, a relatively advanced feature). This prevents directories with a common name, such as string, from unintentionally hiding valid modules that occur later on the module search path. In the simplest case, __init__.py can just be an empty file, but it can also execute initialization code for the package or set the __all__ variable, described later.

Quick template: https://files.catbox.moe/g5tcy5.tar

Also, don’t delete this post. Others might need this information too.

If you have to import a directory structure, you should make each directory a module by creating an __init__.py file in them, and use relative import statements. I usually have one main.py as the entry point, which imports a lib module that contains all of the program logic.

├── lib
│   ├── __init__.py
│   ├── module_content.py
│   └── submodule
│       ├── __init__.py
│       └── submodule_content.py
└── main.py

You can import the lib directory as a module:

main.py:

from lib import some_fn

Within any module, though, you should use relative import statements to import from files and submodules, and regular import statements to import packages from the system or the venv:

lib/__init__.py:

from .module_content import some_fn # import from a file
from .submodule import some_other_fn # import from a submodule directory
from os.path import join # import from an installed package

Items that you define in __init__.py or import into it will be available to import from the module: from .submodule import some_fn. Otherwise, you can import an item from a file by specifying the full path: from .submodule.submodule_content import some_fn.

You can also import an item from a parent package using the .. prefix: from ..some_other_submodule import some_fn.

At work, we use PiSignage for a large overhead screen. It’s based on Debian and uses a fullscreen Firefox running in the labwc compositor. The developer advertises a management server (cloud or self-hosted) to manage multiple connected devices, but it’s completely optional (superfluous in my opinion) and the standalone web UI is perfectly usable.

You can absolutely use it without a reverse proxy. A proxy is just another fancy HTTP client that contacts the server on the original client’s behalf and forwards the response back to it, usually wrapped in HTTPS. A man in the middle that you trust.

All you have to do is expose the desired port(s) to all addresses:

# ...
  - ports:
    - 8080:8080

…and obviously to set the URL environment variables to localhost or whatever address the server uses.

It’s turtles GCC all the way down.

I don’t know which feature you mean, can you link the documentation?

I used it for a while, and it’s a decent solution. Similar to Tailscale’s subnet router, but it always uses a relay and doesn’t do all the UDP black magic. I think it uses TCP to create the tunnel, which might introduce some network latency compared to Tailscale or bare Wireguard.

Right… my mistake, I guess I had SSH config entries in Termux and never questioned whether SSH was using those or DNS.

Still, try to find some way to check which server is being queried. It might reveal connectivity problems with the local DNS server.

Install Termux, then use either the dig or nslookup command to query the DNS name, and check which DNS server is queried. If it’s the private server’s address, you might be having connectivity issues. If it’s 100.100.100.100, the resolver is still trying to query Tailscale’s MagicDNS.

i3 has tabbed windows, and it stands to reason that Sway should have it too.

Hyprland has window groups: https://wiki.hypr.land/Configuring/Dispatchers/#grouped-tabbed-windows

Niri has a feature like that, but a little different since it’s a scrolling tiler. A column that contains two or more windows can be switched to tabbed mode, which displays one window at a time with full height, but you can’t have a tabbed group that is a member of a column, only full tabbed columns.

private dns setting of android

Probably. If that setting is enabled, Android (including Graphene) defaults to 8.8.8.8 if the higher-priority DNS servers (manual or received from DHCP) don’t support DNS-over-TLS or DNS-over-HTTPS.

Proxmox is my number one choice. It’s based on Debian, and has an excellent, extremely straightforward web UI for managing virtual machines and LXC containers.

It’s perfectly reasonable from the perspective of corporate scum: take away a standard feature, then sell it back as an extra. As far as I know, the modem still had UPnP for applications that rely on it.

No, I got it from the horse’s mouth: my WAN address was publicly routable all along, the ISP just disabled those NAT-related features remotely.

I finally got my ISP to enable bridge mode on my modem.

I also learned that I didn’t lose port forwarding and related services because I had been moved behind CGNAT or transitioned to IPv6 – they simply no longer offer port forwarding to residential customers. Ruminate on the implications of that statement so I’m not the only one with blood pressure in the high hundreds.

Even in the open source community, the libre-ness of a product is just one of many factors. The fitness for a purpose, the initial difficulty of the setup, the continuous difficulty of operation and maintenance, the pace of development (if applicable), the professional or community support structure, the projected longevity of the product or service, and the general insanity of the people involved are all important factors that can, and often do outweigh the importance of open software.

Local Unbound with Tailscale’s split DNS has been solid for me. I use it as an OPNsense service with the web GUI, but the standalone YAML config looks simple enough.