Updated 2 February 2026
Licence and attribution
Two licences, because a dataset and a program are different things. The data is CC BY 4.0. The reader is MIT. Neither of them is a warranty.
The dataset: CC BY 4.0
Attribution is the only condition. In exchange you may do essentially anything else, including things people often assume are off limits:
| You may | Condition |
|---|---|
| Copy and redistribute the files, whole or in part | Credit as below |
| Use it commercially, inside a product or a paid report | Credit as below |
| Reshape it - convert, filter, merge with your own data, load it into a database | Credit, and say that you changed it |
| Publish your derived version under a different licence | Credit, and do not imply we endorse it |
| Quote single values in a paper, slide or drawing | Credit, ideally with the release number |
There is no requirement to share your changes back, no restriction on the field of use, and nothing viral: putting CinderCore rows into your own table does not put your table under CC BY.
The attribution string
Copy this verbatim and it is unambiguously enough:
Material property data from the CinderCore dataset v1.6.0 (2026-09-28),
cindercore.dev, licensed CC BY 4.0. Modified: rows filtered to the
magnesia family and converted to metric tonnes.
Drop the last sentence if you did not change anything. In a figure caption, "CinderCore v1.6.0, CC BY 4.0" in small type is accepted practice and we are not going to argue with it. In software, a line in your about screen or licence file is fine; it does not have to appear next to every number.
Why the release number matters
Values change between releases, and some of those changes are corrections of real errors. An attribution that names v1.6.0 lets a reader reproduce exactly what you saw; an attribution that says only "cindercore.dev" sends them to whatever is current, which may now disagree with your figure. Eight insulating densities moved in v1.5.1 alone.
What is actually being licensed
Measured values are facts, and facts are not owned by anyone. If you take four numbers out of this dataset, you are not using our copyright at all - you are using the publisher's numbers, and you should cite the original source, which sources.csv gives you in full.
What the licence does cover is the compilation: the selection of grades, the schema, the normalised units, the family assignment, the confidence flags and the slag index. That is the work this project did, and it is what we ask to be credited for. For anything beyond a handful of values - a whole family, a column, the file itself - the compilation licence is the one that applies.
The reader: MIT
cinderq and the validation schema are MIT licensed. Keep the copyright notice with the source, and do what you like with it, including vendoring it into a closed product. The dataset licence and the tool licence are independent: you may use the tool with your own data, or our data with your own tool, and nothing special follows from either.
What no licence gives you
The dataset is published as is. There is no warranty of accuracy, completeness or fitness for any purpose, and nothing here constitutes engineering advice. A lining that fails is a lining designed by someone, and that someone is not a table of transcribed literature values.
- Values are indicative of a material class. They are not acceptance criteria and they are not a substitute for a datasheet on a delivered batch.
- 61 records carry a low confidence flag, and that flag exists because we expect those numbers to be wrong at the margins.
- Service temperatures assume the atmosphere named in the record. Used against a different atmosphere they can be out by hundreds of degrees, which is why
air_limit_cis a separate field. - If a decision is expensive or dangerous, buy the standard, run the test, or ask the supplier for a certificate. The dataset is a starting point for that conversation, not its conclusion.
The project name
Use the name freely to say where data came from - "based on CinderCore v1.6.0" is exactly what attribution is for. Do not use it to suggest that we checked, certified or endorsed your product, report or service, because we have not. If you are building something that redistributes the data at scale and are unsure how to word it, write to data@cindercore.dev and we will tell you what we would be comfortable with.
Citing a release in a paper
CinderCore (2026). CinderCore refractory property dataset, release 1.6.0.
Sheffield. Available at cindercore.dev. CC BY 4.0.
Accessed 2026-10-12.
Each release ships a SHA256SUMS file, so if you need to prove later which bytes you used, record the checksum of the CSV alongside the citation. Reviewers have asked for this twice and it took thirty seconds both times.