The most reassuring preservation message is also one of the least useful: “Everything is backed up.” Backups answer where a copy might be. They do not explain whether every file was captured, whether its content has changed, which software can interpret it or who is allowed to access it.
A sustainable collection needs at least four kinds of knowledge: what the files are, where they came from, whether they have changed and how they can be used. This guide explains the core technical controls without assuming an enterprise preservation platform.
Start by identifying the actual file format
A file extension such as .pdf, .doc or .jpg is useful metadata, but it is not definitive evidence of a format. A file may have been renamed, exported incorrectly or saved using an older encoding or variant. Identification software can examine signatures and other properties to produce a more dependable inventory.
For reference information, the UK National Archives maintains PRONOM, a registry describing digital file formats and software-related information. Record uncertainty rather than forcing unknown files into an incorrect classification.
For websites, the inventory should include HTML and other generated pages, images, downloadable documents, fonts, scripts, stylesheets and significant API or data dependencies. A capture that retrieves only the homepage is not a complete web preservation record.
What is fixity and why does it matter?
Fixity is the property of remaining unchanged. In practice, organisations commonly calculate a cryptographic digest, or checksum, from a file's bytes. If a SHA-256 digest generated at acquisition matches the digest generated after a transfer or storage check, that supports the conclusion that the byte sequence has not changed.
A checksum does not prove that the original file was correct, that it belongs to a particular person or that it has not been manipulated before acquisition. It demonstrates consistency relative to a recorded reference. Keep a protected copy of the original checksum manifest, the method used and the dates of verification.
Example workflow:
- Receive a file and preserve its original filename and context.
- Calculate SHA-256 and record the exact bytes and acquisition source.
- Copy it to the intended preservation storage.
- Calculate SHA-256 again and compare with the original.
- Record the result, any errors and the actions taken.
- Repeat verification according to the collection's risk and storage policy.
If a checksum changes unexpectedly, do not immediately overwrite the preserved copy. Investigate which version is authoritative, retain the evidence and document any repair.
Metadata makes a file intelligible
Useful metadata may include title, creator, creation or publication date, source URL, date collected, rights, format, checksum, software dependencies and relationships to other material. Technical metadata is not interchangeable with descriptive metadata. Both may be needed.
For a historic website, provenance should distinguish original content from reconstructed content. A screenshot from a web archive, a source document supplied by the rights holder and a newly written historical summary should not be presented as the same kind of evidence.
When a PDF cannot lawfully be hosted, a bibliographic record can still capture its author, title, year, institutional repository and access route. This is how our research library approaches papers previously linked from the Digital Curation Exchange.
Web archives, backups and restored sites have different purposes
An archive snapshot documents a representation of a site at a particular point in time. A backup may contain source code, databases and configuration that are needed to run the application. A restored site is a functioning publication created from what can lawfully and technically be recovered.
They complement each other. A public archive may contain the text of a lost page but none of its CMS configuration. A database backup may contain all the text but not a record of how visitors actually experienced it. Combining evidence improves the result, provided the relationship between sources is documented.
See website archiving for capture planning and website restoration for recovery after a loss.
Do not overlook the URL inventory
A URL is part of the documentary record. Research papers, old support tickets and university reading lists may cite a specific file or node identifier. When a site changes, preserving that reference can be as important as preserving the homepage.
A good migration records the original URL, its previous subject, the best new destination, status code and verification result. A permanent redirect should lead to substantially equivalent content. If there is no suitable successor, an honest missing-resource response is better than sending readers to an unrelated sales page. Learn more about website migration and SEO recovery.
Questions about formats and integrity
Is SHA-256 enough to preserve a file?
No. It helps verify byte-level consistency, but it does not prevent storage loss, document rights, provide access software or record the object's meaning.
Can a file be intact but unusable?
Yes. Its bytes may remain identical even when the necessary hardware, software, credentials or contextual documentation are no longer available.
Should preservation copies be converted into modern formats?
Conversion may improve accessibility, but it should be documented and assessed. Retaining the original alongside a derived access copy is often useful when lawful and feasible.
What if the original file cannot be found?
Record its citation, known location and the attempts made to retrieve it. Do not recreate its apparent findings from a filename or claim an unrelated document is a replacement.
Related resources
- The National Archives: PRONOM
- COPTR preservation tools and workflows
- Digital forensics in collecting institutions
- Digital preservation hub