Open Music Data and Licensing — Doing Datasets Right
A dataset is a collection of works — its licence answers four questions
Related
A MIDI dataset, a folk-song archive, a stack of sample packs — a dataset's licence is a different question from a single song's. It is a licence for a *basket of works*, and you need a sheet that explains the whole basket.
(Rules and practice; not legal advice.)
Four questions a dataset licence must answer
1. Commercial use? — datasets are often assembled from many sources, and each block may answer differently (this project's C1/C2/C3 are three answers); 2. Redistribution? — may the dataset (or a subset) be re-released, as-is or repackaged; 3. Attribution? — to whom, and how (dataset chains are long: original source → compiler → dataset); 4. Derivatives? — may users adapt, remix, rework.
A proper dataset notice answers these four, line by line — not a vague "for study use".
The common licence families for data and content
- Creative Commons (see the CC family) — the most general;
- CC0 / public-domain marks — maximum freedom;
- The ODC family — licences tailored to databases (BY, ODbL, etc.);
- Custom terms — common in institutional archives ("study and research only" is the classic tier).
Custom terms do not mean un-open — they narrow freedoms to specific uses. Read the terms and follow them.
A transparent practice: this project
This project's dataset is licensed per track, by tier:
- C1 (commercial): free to use, including commercially;
- C2 (non-commercial): non-commercial use;
- C3 (study/research only): restricted use;
- Soundfonts have their own F1–F4: from "free to distribute" to "not listed".
Every record carries its tier; every example on the pages shows source and tier — read the tier before reading the data is this project's default move.
Three moves for users
1. Before downloading: read the licence (find the four answers); 2. While using: respect the tiers (commercial need → C1-class material only); 3. When publishing: keep the attribution chain (every link from original author to dataset).
Listen: one specimen in an archive
A dataset is countless "specimens" — hear a folk song walking out of an archive, carrying every label: source, era, compiler, licence.
Next
- The general licence family: the CC licence family.
- Work and recording layers inside a dataset: how many rights live in one song.
- The smallest unit in any dataset: Melody.
Listen in the library
All tracks from midicn-lib — inline badge is the license tier (C1 commercial / C2 non-commercial / C3 study only).
Sources
- 本文为数据集许可的通识介绍，并以本站数据集的分级实践为例；不构成法律意见
- 涉及的曲目均标注来源与许可档位，可在线试听
Used for fact-checking only; all prose is original.