What programming/technical projects have you been working on?

I'm going to be installing Proxmox on my server PC (it has been transferring data back and forth for weeks while fixing a ZFS issue), and I figure who better to ask about architectural decisions than other techies on Tildes.

Everything (Wireguard, Nginx, Forgejo, Jellyfin, Samba, various game servers, etc) is currently run inside a single Linux instance that is running on baremetal. I enjoy the simplicity of this solution, but I should learn to do things in a more industry-standard way someday.

I know that the final configuration will look like one Proxmox instance running a number of programs across a number of Linux VMs (or containers?), with an additional Windows VM that will allow me to experiment with SR-IOV and Sunshine/Moonlight based game streaming (yes, I understand that the performance will not be stellar). However, I'm lost as to best practices when it comes to separation of concerns.

Do I want a singular VM, which all networking is passed through to, which runs Nginx and acts as a reverse proxy for other VMs (this will also allow me to insert "system under maintenance" messages when certain services are down)?

Do I want to pass the primary ZFS array through to a VM which handles dissemination of access to the array contents using NFS or similar? How does this consideration change if I only care about Linux VMs accessing said files? Should this VM be a different one to that which performs network proxying, so as to sandbox the network stack?

How much should I run on the Proxmox instance itself, vs inside the VMs? I'm completely stuck in the choice paralysis stage, and I desperately wish to avoid over-complicating the final solution.

As is tradition, as soon as I finish up work for the year, my brain goes "you should be writing code". It's been nearly a year since I released my Tiny Flowers mod for Minecraft, and it has one major issue that I knew about from the start that I want to change.

The issue is the block state explosion. The game has two ways of storing data about blocks, the simplest of which are block states. My little garden blocks had 4 slots that could have 20+ different values in each and the block could be rotated one of 4 ways, and the game had to enumerate and build block models for each of these combinations. Also, these combinations were constant and couldn't be expanded upon by other mods.

My plan with V2 of the mod was to use a block entity instead, which is an entirely different system for storing data about a block, and use a data pack to define all of the flower types. The latter point would mean that other mods could define their own tiny flower variants and it would just work. I've done most of the rewrite now, I just have the hardest part left: rendering.

That system of enumerating every possibility was extremely convenient to use, as it took advantage of an existing system in the game, but now I don't really know what to do. There are pages on the Fabric wiki and docs (Fabric being the mod loader I'm using), but I don't know enough yet to know what I need to know from them. It also seems like the way the game wants you to do custom rendering is changing at the same time, so I don't even know what the best practice is here.

Ideally, I'd have something similar to the block state system but on the fly, where I can go "I want this model, this model, and this model all together with these textures" and use existing game code. I don't know if that exists or how to find it if it does, but that's what these kinds of projects are all about: doing something you haven't done before to learn how to do it.

Not a particular project but I'm going to do a little bit of inter-business consulting. My company has sister companies that also do e-commerce and one of the businesses seems to be struggling with their backend so I'm looking over a Rails and React app (both things I have little experience with) to figure out where we can make some improvements.

For myself, I am thinking of working on a small program that lets me manage clients, jobs, and generate invoices for some of the freelance audio work that I do. I have following Gleam a lot and I think this is a great tool for building the application.

Ever since switching to Helix from Jetbrains IDEs as my editor/s of choice earlier this year, I've been obsessed with going CLI-first as much as possible for my personal projects. I'm now working on moving my finances to text with beancount.

The biggest challenge for me with this project was actually deciding whether to go with beancount or with hledger. I do not feel qualified to write about the tradeoff, except that I prefer the syntax of hledger and I prefer some of the clear decisions the beancount author made in this post. That post is a bit outdated, though, and I have not spent enough time comparing the two pieces of software to say if my opinion is "right" (I may switch!).

I made the call that I did because beancount is clearly much more opinionated than hledger, which tends to work better for me from a meta-perspective. If a piece of software gives me too much freedom, I tend to spend time messing with it to get it to behave in a certain way. A piece of software designed to force me to do things a specific way tends to push me to get the actual project at hand done. I also suppose that I can transition between the two without too much difficulty.

Also, I'm not sure what the best way to store this plain text file securely is. Naturally, it's got a decent amount of sensitive data on here that I try to lock behind some kind of authentication. I am not a security engineer; I just know how to code. I could encrypt it, decrypt it, and re-encrypt it every time I was to use or update it, but if anyone here has better suggestions, I would love to hear them.

Briefly mentioned this in my comment in the weekly post on ~talk but I'd love some input on the following.

I've currently got a little HP office PC running Debian as my home media server. It's got 2x4TB HDDs in a software RAID 1 through mdadm. I've got Jellyfin and the *arr stack running through Docker, with qbittorrent going through a VPN, all setup with https://yams.media. It's got 3.6TB usable space and I've used up about 2.5TB, and that is filling fast. I'd love to upgrade to get like 15-20TB of space so that I could also start hosting an Immich instance. I wanted to ask here if anyone had any recommendations on what's the best path to upgrade. I was thinking of just setting up a separate RAID 5 or something with 3x12TB drives and start using that for my Jellyfin/*arr stack and then repurpose the 4TB RAID 1 for Immich. Another option I was thinking of was setting up UnRAID to have one big RAID volume and then just having separate subfolders for Immich, Jellyfin, and anything else in the future. I was originally leaning towards the former since that sounded simpler in my head for some reason but I'm tempted by the flexibility and ease of use of UnRAID. I'd be ditching the HP in either case as it was meant to be a temporary/starter server and it doesn't have enough SATA ports. What would you guys recommend?

I am also looking for VPN recommendations. I've currently got Private Internet Access since I got a deal for 3 years for $80US. However, it's been almost 3 years now so I'm looking for something new. I'm like 99% set on getting Mullvad but am curious if anyone recommends anything else.

I recently made a donation portal for a non profit. Should go live soon! The whole thing is stateless (well, all state is in query parameters or Stripe). It’s basically a Stripe front end. There’s only about 50 lines of client JavaScript.