This is revision 2 of this page, effective September 25, 2026. Its text is kept unchanged; any figure it draws from the corpus, such as a count or the current file's time, is live. The current revision is in force.

Corpus

The whole archive as one file: every message, its date, its structured fields, its composition score and typing figures with the method version, its set and position, the hash of its message, the revision of the terms it was published under, and a tombstone record for anything removed under the published procedure. JSON Lines, one record per line, UTF-8, with a header record first that carries the schema inline and describes the file well enough to read it with nothing else. The same records are also issued as one plain UTF-8 text file, in chronological order, each entry preceded by its date in ISO form and in words, its position, and its labelled fields; plain text is the format with the longest demonstrated life, and the one a reader with no tooling can open.

The file is regenerated on every publish. Dated snapshots sit beside it with checksums, so that a citation points at something that does not move. Everything here is in the public domain under CC0 1.0; see how to cite.

Checkable, not immutable

The archive does not ask to be believed when it says nothing has been edited. It says "checkable", and this is the check. Every entry carries the SHA-256 of its message, which survives into its tombstone. Snapshots are taken on the first day of January, April, July, and October, are written once and never replaced, and each one's header names the previous snapshot and its hash, so the series is a chain. Each snapshot is deposited with custodians outside the operator's control as they are added below, and a Wayback capture of this page and of the snapshot file is requested at each one. The command python -m mwcn.cli verify OLDER NEWER in the public repository confirms that every record in an older file is present and unchanged in a newer one, allowing only a set completing or a message becoming a tombstone with its hash intact, and that the newer file names the older by hash. Once a year the site is rebuilt from a snapshot on a clean machine and the result recorded in the log.

Dedication of the compilation

Each message is in the public domain by its writer's own act. Everything the archive adds around the messages (the schema, the ids, the timestamps, the fields, the figures, the tombstones, the snapshots, and the arrangement of the whole) is dedicated to the public domain under CC0 1.0 by the archive, with an express waiver of any sui generis database right or similar right in the compilation, wherever such a right exists.

Datasheet

A datasheet answers the questions a careful user of this data should ask before using it: who wrote it and who is missing, how the first writers were found, what the score can and cannot show, what re-identification risk a signed message carries, and what the corpus cannot tell you. The Croissant metadata and the dataset card are derived from the same facts.

Revision 2, effective September 25, 2026.