How to Cite an Open Dataset (Even When the Licence Doesn't Require It)

Every dataset on this site is CC0. You can copy it, ship it in a commercial product, and never mention where it came from. Nothing in the licence obliges you to credit anyone. You should still cite it, and not out of politeness.

A citation is a debugging tool

Six months from now a number in your output will look wrong. Someone will ask where it came from. If the answer is "a JSON file we downloaded once", the investigation stops there. If the answer is a URL and a date, the investigation continues: you fetch the source again, compare, and find out whether the data changed or your code did.

That is the whole argument. Attribution is a side effect.

What actually needs to be recorded

• The dataset name and its permanent URL. A link you can re-fetch, not a local filename.
• The version or retrieval date. Reference data changes quietly. "As retrieved on 12 September 2026" is enough when no version number exists.
• The publisher. A person or an organisation. In that order if it is one person.
• The licence. CC0 here, but record whatever it is. Licence questions arrive later, usually from someone else's legal team.

That is four fields. Anything more elaborate tends to go stale before it gets used.

A format that survives copy-paste

Academic styles assume a paper. For reference data, a plain sentence works better because it survives being pasted into a source comment:

Programming Languages dataset, YJTOON. Retrieved 12 September 2026 from https://yjtoon.com/catalog/programming-languages/ (CC0).

If you need something closer to a formal style, the pattern below is easy to generate and easy to read:

Vurgun, Y. E. (2026). Programming Languages [Data set].
YJTOON. https://yjtoon.com/catalog/programming-languages/

Put it where it will be found

A citation in a README is better than nothing. A citation next to the data is better than that. In practice there are three places worth the effort:

1. In the code that fetches the data. A comment above the fetch call with the URL and the retrieval date.
2. In the output, if the output is user-facing. One line, not a paragraph.
3. In the prompt, if the data goes into a model. A source line in the system prompt reduces invented detail, because the model has a specific string to be consistent with. The reference card template shows the shape.

When the licence does require attribution

CC0 waives attribution. CC BY does not. If you are pulling from a source licensed CC BY, ODbL, or a custom data licence, read the actual terms rather than assuming. The differences matter for commercial work, and the distinctions are worth understanding before someone asks — what CC0 actually covers walks through them, and attribution practice covers the mechanics of crediting correctly.

What to do about derived datasets

If you filter, merge, or reformat the data, you have made a derived work. Say so. "Derived from the YJTOON Networking dataset, retrieved 12 September 2026" is honest and takes six words. It also warns the next reader that the rows they are looking at may not match the published source.

That warning is worth more than the attribution itself.

Automate the boring part

Every page in the catalog prints a ready-made citation with the dataset name, year, and permanent URL, so there is nothing to compose. The remaining work is discipline: put it in the commit that fetches the data, not in a note that never gets read.