I'm not an expert on btrfs, but I assume the inconsistencies come from deduplication, metadata, and maybe compression. I think some of them just count raw block storage, and some include the cost of metadata.

Traditional du assumes that each file takes up it's full space on disk which isn't always the case on btrfs. When using btrfs backed oci images, storage can easily appear multiple times higher.

I use btrfs filesystem usage /. I'm not sure that it is the "correct" way, but it works fairly well.

Standard forgejo shoutout. It is a fork of gitea with more features following the foss philosophy. It is codeberg's backend https://forgejo.org/2024-02-monthly-update/

You can still compile infinity from source with your own api key

1. Codeberg is fully open source(forgejo) while gitlab has an open source core+community edition but a source available propietary enterprize edition.

2. Codeberg is a nonprofit with no ulterior motives. Gitlab is a publicly traded for profit entity with a goal to make profit

3. This could just be me, but codeberg feels a lot more transparent. When they have outages, they explain why.

4. Super minor, but the codeberg team "self-hosts" their own servers so you only need to trust the one entity rather than additionally trusting the server provider.

Primary code editor: helix

Graphical debugger and certain IDE features: vscodium

Lots of open source language servers: clangd, rust-analyzer, perl-navigator, ...

Makefile to compile-comands.json: bear

TUI file manager: yazi

Better Grep:ripgrep

Debugger: gdb(gnu debugger)

Section 4 is what gets me. Your rights are temporary and revokable meaning the the rest of the license doesn't matter in the long term

## Section 4: Termination, suspension and variation
1. We may suspend, terminate or vary the terms of this license and any access to the code at any time, without notice, for any reason or no reason, in respect of any licensee, group of licensees or all licensees including as may be applicable any sub-licensees.

Zellij - a better way for a cli application to communicate with the terminal

Warp - a terminal emulater that integrates LLM completion natively

Fish - a shell that generates completions automatically from a man-page

Now there are 3 competing standards Edit: 6ish accually

Turing Complete Configuration

• more extensible
• tend to be heavier
• harder to provide detailed error messages
• more difficult for new users

Data Based Configuration

• easier to use
• easier to provide documentation
• lighter to embed
• more limited usecases

Nixos

Pros

• Delarative Config
• largest package repos

Cons

• poor documentation
• cli and package management is in limbo with unstable flakes/cli

In most cases you don't want one. It can make forks confusing and lend malicious actors more credibility than they deserve.

Copyright controls the code. Trademarks are the recognisable names/icons that identify a project.

I use a shared boot partition all the time. I mount my EFI system partion on /efi. Then I bind mount /efi/$OSNAME to /boot in my fstab. Then I just manage my bootloader (typically systemd-boot or refind) manually. Any distros I install are installed in my encrypted btrfs partition within their respected subvolumes