101 lines
4.8 KiB
Markdown
101 lines
4.8 KiB
Markdown
---
|
|
title: "Day 6 Updates"
|
|
description: "I don't really have much to say today so here are just a few things I've been working on"
|
|
author: Amolith
|
|
cover: /assets/pngs/calendar.png
|
|
date: 2020-05-03T01:57:03-04:00
|
|
draft: false
|
|
toc: true
|
|
categories:
|
|
- Technology
|
|
tags:
|
|
- Zettelkasten
|
|
- Doom Emacs
|
|
- Emacs
|
|
- NixNet
|
|
- 100 Days To Offload
|
|
---
|
|
|
|
I haven't been able to come up with a specific topic for today so this
|
|
is just a kind of generic update about me.
|
|
|
|
## Zettelkasten
|
|
In my [previous post about Vim,](/vim-as-a-markdown-editor/) I briefly
|
|
mentioned being inspired to create a Zettelkasten by Daryl Sun in [his
|
|
fourth 100 Days To Offload
|
|
post.](https://write.privacytools.io/darylsun/100-days-to-offload-day-4)
|
|
A Zettelkasten is a personal knowledge management tool that allows one
|
|
to quickly retrieve useful information about a subject, relearn
|
|
forgotten concepts, and discover connections between those concepts to
|
|
form entirely new ideas. There are different processes recommended by
|
|
different people but I think it's a very personal choice and depends on
|
|
what your workflow will look like. Mine will be as follows.
|
|
|
|
1. Take *very* concise notes on something I learned in a *physical*
|
|
notebook
|
|
2. When I'm able, go through those notes and add them to my [digital
|
|
Zettelkasten,](https://git.nixnet.xyz/Amolith/zettelkasten) expanding
|
|
them a little and fleshing the thought out more
|
|
|
|
The last step is *the most important* as this is the one where you sit
|
|
down and think about what you're adding and try to draw connections
|
|
between it and what you already know. The goal is not to make the
|
|
longest and most complete notes in the world but to add value to each
|
|
*concise* thought by linking it with others and build a web for you to
|
|
explore later. You might not see immediate benefits but a mature
|
|
Zettelkasten with hundreds of entries will constantly surprise you as
|
|
you tumble into your own store of knowledge and rediscover things. That
|
|
surprise is actually one of the greatest benefits to this kind of
|
|
knowledge management system; when something is surprising, we tend to
|
|
remember it better.
|
|
|
|
## Doom Emacs
|
|
A friend of mine convinced my to try [Doom
|
|
Emacs](https://github.com/hlissner/doom-emacs) and, so far, I am very
|
|
impressed. Emacs itself is very powerful but, from what I can tell, this
|
|
configuration adds a *lot* of value. The main one being Vim keybindings
|
|
:wink: I'm looking forward to learning [org-mode](https://orgmode.org/) and
|
|
seeing what it can do for my productivity. As a text editor and
|
|
programming tool, I plan to stick with [Neovim](https://neovim.io/) on
|
|
desktop/laptop, [Vim](https://www.vim.org/) on Debian-based systems,
|
|
[vi](https://wikipedia.org/wiki/Vi) wherever else.
|
|
|
|
## NixNet plans
|
|
Today, I fleshed out some of my thoughts on reprovisioning all of my
|
|
over the summer. I'm going to have [Ansible](https://docs.ansible.com/)
|
|
or [Salt](https://docs.saltstack.com/en/latest/) build and deploy
|
|
[LXC](https://linuxcontainers.org/lxc/introduction/) containers to a
|
|
baremetal server from [Hetzner](https://www.hetzner.com/sb) running a
|
|
*very* minimal [Alpine Linux](https://alpinelinux.org/) installation.
|
|
Whatever setup I have for those will of course be available on
|
|
[Gitea.](https://git.nixnet.xyz/NixNet) From there, my local NAS will
|
|
use something like [borgmatic](https://torsion.org/borgmatic/) to back
|
|
up files and databases from all of my servers and
|
|
[LXD](https://linuxcontainers.org/lxd/introduction/) to create container
|
|
snapshots[^1]. All of that will be mirrored to
|
|
[BackBlaze](https://www.backblaze.com/) likely using their B2 model as
|
|
paying per GB per month is generally the most reliable option. Under one
|
|
of the others, there's always the possibility that I might upload more
|
|
than they think is reasonable and start limiting me in some way.
|
|
|
|
Short-term, I'm going to consolidate some of my servers to a single
|
|
baremetal machine from Hetzner. Long-term, I'm going to look into
|
|
building and racking my own servers in a datacenter in Germany, likely
|
|
one of Hetzner's. This comes with a plethora of benefits but a pretty
|
|
major detriment: the up-front cost will be absolutely *massive*.
|
|
Building a rack server worth putting in a datacenter will be incredibly
|
|
expensive at the start. Following that, all I have to pay is a monthly
|
|
fee for however much space it uses in the rack and it won't be too much.
|
|
Before any of that is even considered, I'm going to be spending a lot of
|
|
time discussing things with my father; he did a lot of racking before he
|
|
got his current sysadmin job and has a ton of advice to give, from using
|
|
VoIP to powercycle the server to what networking gear to look at and how
|
|
to organise everything within the rack.
|
|
|
|
I have a lot of really big plans.
|
|
|
|
[^1]: This one isn't *really* necessary as building the containers with
|
|
Ansible/Salt is automated and it's a simple process to rebuild them.
|
|
Snapshots might just take less time to redeploy should something go
|
|
wrong.
|