Open Data

The Mangrove dataset is open data. Anyone can download all of it, reuse it under the license of each review, and host a copy of it.

Getting all the data

  • api.mangrove.reviews/dump returns the whole dataset, and a new dump replaces it every night. The Accept header selects the format: JSON by default, text/csv for CSV, and application/n-triples for RDF.
  • The API returns reviews by subject, reviewer and time. With gt_iat it returns the reviews issued after a given Unix time, newest first, so a copy fetches only the reviews it lacks. A copy pages through a large update with offset until a response holds fewer reviews than limit.

Both are open to anyone, with no account or key needed.

License

Each review is licensed under CC-BY-4.0, or under CC-BY-SA-4.0 where its license field says so. A reuse must credit the reviews and show where they were changed, as the Terms of Service set out. A reuse must also respect the privacy of reviewers and copyright law.

Every review carries its proof

Each review is a JSON Web Token signed with its reviewer’s key, and the JSON dump holds the token of every review. Anyone can check each review of a copy against the public key in its kid, so the signature of each review vouches for it, rather than the server that delivered the copy. Two copies merge by their signatures, since a signature identifies a review.

A reviewer can delete their review, and the operators remove reviews that break the Terms of Service. The next dump omits both kinds, and a copy drops them by comparing its signatures with that dump.

One dataset, many copies

The Open Reviews Association runs one server, and every integration writes to it and reads from it. One logical dataset lets a review written in one app appear in every other app. The association also imports reviews from other open sources into it, so that everyone can reach them in one place.

Copies and exchange are welcome. The Mangrove Review Standard makes the dataset easy to sync. The server code is open source under the Apache License 2.0, so anyone can host a copy that serves the same API. As more apps and organisations take part, the dataset can spread over several servers that replicate each other. If you are interested in hosting a copy or exchanging reviews, write to us on Matrix.