Pull Requests

Every change to the dataset goes through a pull request. The pull request is what makes a change visible and reversible.

What happens after you open one

Automated validation runs the same checks as npm run validate, plus the tests and a reproducibility check. It has no access to secrets, because it runs on code that anyone can submit.
Annotations appear on the lines that caused a problem, so you can fix them without reading a log.
A maintainer reviews the content, the rights statements and the design of the change.
Merge moves the change into main.
Publication builds the catalog and uploads it. Publication runs only on main, in a protected environment that holds the storage credentials. The upload is checked on the public host before the catalog is replaced.

The pull request template asks for:

the server's name, id, addresses, website and Discord;
the owner, meaning the person or team responsible for the listing;
the type of change: a new server, an update, a removal, or tooling;
confirmations: that you have permission, that the information is accurate, that the media follows the specifications, that no copyrighted or trademarked material is used without permission, and that you understand the licence terms.

The confirmations are a statement of intent, not a legal guarantee. They make the responsibility explicit.

Say what the server is, in one paragraph, without marketing claims you cannot support.
Use the server's real addresses and only the Minecraft versions it actually supports.
In the notes, explain anything a reviewer must know: an unusual wildcard, a change of brand, a provenance question.

An update is a pull request that changes the files under servers/<id>/. Keep the id.

A removal deletes the directory and adds the id to retired-ids.json, with a reason. The id is retired permanently. See server ids.