M³: Minecraft Mod Machine

Discord

Modified from xkcd 918, licensed under CC BY-NC.

This is a project to build a repository/package manager for Minecraft mods, allowing modders to easily publish mods, players to easily download mods, and modpack creators to easily, ahem, create modpacks. This project is currently in its earliest stages. Once we get a rough idea of exactly what we want in such a tool and what parts would be necessary, we will get started on coding.

Wishlist

These are the features we would like to have. For now, anything goes. Once the list has pretty much settled down, we will separate it into phases so we can start implementation. Have an idea? Send a PR and we'll add it to this list.

• Easy mod installation
  • This should be as simple as searching for a mod's name and clicking install.
  • Dependencies should be automatically installed as well.
• Easy modpack assembly
  • Users should be able to easily distribute a list of mods along with any additional files (e.g. configs) for other players to install.
  • Versions should be kept identical, with a possible exception for versions which are marked as exclusively bug fixes.
  • Modpacks should be able to be forked and re-distributed easily.
• Optional accounts
  • The only accounts people should need are Mojang accounts. If they wish to share mods or modpacks, they may need a seperate account.
  • If possible, it would be good to emulate Minecraft's own auth flow in our client so that Mojang accounts could be the only accounts necessary at all.
• Compensation
  • To gain any traction among modders, we need to be able to compensate modders somehow. Ideas for how we could do this:
    • An option within the client where players could donate to an individual modder or make a donation to be equally spread among the authors of all mods in an instance.
    • Light ads in the client
• Lightweight client
  • During gameplay, the client should use the smallest possible amound of system resources while still being useful to allow more resources to be dedicated to the game. This probably means not keeping all game logs in memory and instead loading old logs automatically as the user scrolls up, or just writing them to a file.
• CI service for modders
  • In the "winning modders over" department, we'll want to do something BETTER than the current leader, Curse, in order to succeed. A CI service would be one way to do that. We could automatically watch GH repositories for new tags, and when they appear automatically compile them and make the new release available for download. This means that, aside from a few minutes of initial setup, adding a mod to MMR would be a "set it and forget it" thing, meaning that adding a mod to MMR becomes less of a "why" and more of a "why not."
  • We will also want plugins for existing CI systems for people who alreafy hsve that set up.
• Simple server setup
  • Allow a player to enter a server IP and automatically get an instance with that server's modpack and the server's IP preloaded into the server list.
• Single local mod repository
  • Forge allows us to store mods in a single local repository instead of in the mods folder of each instance. Let's use this to save on disk space.
• Resource pack support
  • Sort of like the FTB Launcher. It would be even better to independently store the core resource pack and any mod patches for each pack, allowing automatic generation of resource packs for brand new and/or small packs that wouldn't normally get resource packs put together.
• Cross-Platform
  • The client should work on Windows, OS X and Linux from Day 1.
• Customizable Modpacks
  • In addition to the ability to make arbritary customizations to a a pack, it would be good to have preconfigured options to allow players to customize. See ATLauncher.
  • In addition, we should allow including/excluding files (e.g. MineTweaker scripts) depending on installed mods.
• External Mods
  • Modpack developers should be able to include in their packs Mods that have a license that permits modpack inclusion, but whose authors have not uploaded them to our repository. These should proably be handled the same way Curse does: store them on our side as JARs, within the modpack along with configs.
• API
  • Anyone should be able to build a client or other software that uses our data.
• Permissions
  • Modders should be able to privately "release" their mods for a private beta, or publicly release them. If a mod is publicly released, there should be four options: "do not allow in modpacks," "allow in private modpacks", "allow in public modpacks," and "allow in monetized public modpacks." Obviously, even if this last option was selected both the modder and the modpacker would be paid. If the mod was still in a private beta, only the first two options would be available.
  • Modpack developers should be able to make a modpack private (whitelist of player names), or public.
• Launcher Agnostic
  • Besides a built-in launcher, we should have first clas support for MultiMC and the vanilla launcher.
• Maven
  • Provide a Maven for modders to easily get copies of mods to develop with.
• Sandboxed Client
  • Launch MC under an OS-level sandbox (e.g. LXC, OSX sandbox, etc) to reduce potential damage from rogue mods
• Offline support
  • Vanilla supports offline play. We should too.

Architecture (Technical Details)

These are the big components we forsee needing.

Backend - Implementation Thread

This is the system that hosts mod files along with mod metadata and modpacks. The CI system will also be in this component.

Exposed Interface: JSON API

Requirements: Light on resources (disk space, bandwidth, CPU) so that it is affordable.

Client Core - Implementation Thread

This handles storing mods locally, communicating with the backend, and interfacing with Minecraft.

Builds on: Backend

Exposed Interface: Library

Requirements: Light on resourses (CPU, Memory) so that it can run alongside Minecraft without slowing it down

Client User Interface - Implementation Thread (CLI, GUI thread will come later)

This is what the user sees. It needs to allow building modpacks, launching instances, etc.

Builds on: Client Core

Exposed Interface: CLI first, seperate GUI implementation later

Requirements: Easy to use, light on resources (see client core)

Web Interface

This is similar to Curse's CurseForge. It allows casual browsing of available mods, and some of the things the client UI allows (everything but managing instances).

Builds on: Backend

Exposed Interface: Website

Requirements: Simple to navigate