Direct answer
A canonicalisation package brings the identity and distribution family of one release into a single statement. Record the work or version DOI, canonical URL, title, author, version, date, file name, size, hash algorithm and value, licence, archive mirrors and relationship types. Every platform copy carries the correct relationship to the canonical work. A DOI or number of platforms is not evidence of quality, and a file hash shows only whether the compared file bytes changed.
In plain language
Imagine that a book has a passport. The DOI is its persistent identity number, the canonical URL its official address, the file hash its seal impression, the licence its permission for use, and the archive its preserved copy. A passport does not prove that the book is good; it helps identify which work and which copy you have.
Why this matters
If different files are published under the same name, people and AI systems cannot tell which version is official. An incorrect DOI relationship, outdated file or unclear licence breaks the citation and reuse chain.
Do not confuse
- A work DOI may represent a work family containing several versions.
- A version DOI identifies a particular immutable release.
- The canonical web edition is the publisher's current official human and machine surface.
- An archive mirror preserves a copy; it need not be the canonical publication or independent verification.
- A distribution copy supports discovery; a derived dataset may be a separate resource produced from the work.
What should you do?
- Lock the human and machine files for release and record their names, sizes, media types and hashes.
- Link the work DOI, version DOI and canonical URL reciprocally with the correct relationship types in metadata.
- Align the meaning of title, author, date, version, language, licence and description across all platforms.
- Classify each mirror explicitly as canonical publication, version, archive mirror, distribution copy or derived dataset.
- State which files and elements the licence covers, any third-party exceptions and the citation form.
- After download, recalculate hashes to test bit-level parity of archive and distribution copies.
- Maintain periodic verification records for DOI resolution, canonical URL, file access, metadata and hash parity.
How do you audit it?
- Does the DOI resolve to the correct work and version?
- Are work, version, predecessor-successor and derived-resource relationships typed correctly?
- Is the canonical URL explicit and consistent across platform metadata?
- Do the hash algorithm, value, file name and size match?
- Are the licence scope and third-party exceptions intelligible?
- Was an archive or distribution copy presented as independent verification?
- Were DOI, download or platform counts used as evidence of academic impact or adoption?
Limit
A DOI does not prove authorship, peer review, accuracy or global adoption by itself. A file hash compares copies; it does not establish that a file is correct, harmless or lawful. A licence can be granted only by a rightsholder or authorised party, and the particular choice may have legal consequences.
Remember in one sentence
Identity, address, integrity, permission and relationship are different things; together they form a verification chain.
Sources for this record
- S14DataCite, *Versioning*Identity infrastructure
- S15DataCite, *Connecting Versions with Related Identifiers*Identity infrastructure
- S35W3C, *PROV-O: The PROV Ontology*Standard
- S55DOI Foundation, *DOI Handbook*Official persistent-identifier handbook
- S56NIST FIPS 180-4, *Secure Hash Standard*Official standard
- S58Zenodo, *About Records*Official repository documentation
- S59Creative Commons, *About CC Licenses* and License ChooserOfficial licence-provider documentation

