DIGITAL ARCHIVES · BORN-DIGITAL RECORDS · FIXITY · METADATA · PRESERVATION · ACCESS

How Digital Archives Work

A digital archive does not preserve files by putting them on a hard drive and leaving them alone. It preserves the ability to identify, verify, interpret, render, relate and retrieve digital records across technological change.

A digital record survives only when its bits, meaning, identity, context and usable behaviour survive together.

That is harder than it sounds. Paper can sometimes remain legible for centuries without electricity. A digital file may become unreadable because the storage medium fails, the file format becomes obsolete, software disappears, encryption keys are lost, dependencies break, metadata is separated, a database schema vanishes, a website changes, or a cloud service closes. Digital preservation is therefore active stewardship.

The problem is not simply technological. A file can remain perfectly intact and still lose meaning. A spreadsheet without formulas, a database without its schema, an email without headers, a photograph without capture metadata, or a web page without linked assets may preserve bytes while losing evidence.

The short answer

CREATE / RECEIVE
  → IDENTIFY
  → APPRAISE
  → TRANSFER
  → INGEST
  → VIRUS / SECURITY CHECK
  → FIXITY
  → FORMAT IDENTIFICATION
  → METADATA CAPTURE
  → PRESERVATION STORAGE
  → REDUNDANCY
  → MONITOR RISKS
  → MIGRATE / EMULATE / NORMALISE WHEN NEEDED
  → CREATE ACCESS COPIES
  → AUDIT
  → PRESERVE PROVENANCE
  → FUTURE ACCESS

A digital archive therefore operates like a continuity engine. It repeatedly asks whether the record still exists, whether it is still the same record, whether it can still be rendered, whether its context is intact, and whether future users can understand what happened to it.

1. Digital archive is not digital storage

Storage holds bytes. Archiving governs the long-term survival and meaning of those bytes.

A folder on a laptop is storage. A replicated cloud bucket is storage. A backup is storage plus recovery capability. A digital archive adds identity, provenance, metadata, fixity, format knowledge, preservation actions, access controls, audit history and succession.

2. Born-digital and digitised are different

Born-digital material originates in digital form: email, word-processing files, spreadsheets, databases, websites, photographs from digital cameras, source code, audio, video, chat logs and system exports.

Digitised material begins as a physical or analogue object and is converted into a digital surrogate: a scanned manuscript, photographed map, digitised tape recording or converted film.

The archival risks overlap but are not identical. A scanned book can return to the physical original if needed. A born-digital email may have no analogue original at all.

3. Preservation starts before transfer

Digital records are easiest to preserve when creators use durable formats, stable identifiers, meaningful metadata and exportable systems from the beginning. Poor record creation pushes avoidable complexity downstream.

The United States National Archives and Records Administration, for example, publishes transfer and metadata guidance for permanent electronic records. The principle is universal: preservation quality improves when recordkeeping and archiving are connected across the lifecycle.

4. Appraisal still matters in the digital world

Cheap storage created the illusion that digital archives could keep everything. But preservation cost is not just disk space. Every retained object carries costs for metadata, security, integrity checking, migration, discovery, privacy, rights, format analysis and future access.

Appraisal therefore remains essential. The archive must know what deserves permanent preservation, what has temporary operational value, what duplicates another object, what contains unnecessary personal data, and what should be disposed of according to authorised retention rules.

5. Transfer packages need boundaries

A transfer should not arrive as an unexplained pile of files. The archive needs a defined package: what records are included, where they came from, who transferred them, what restrictions apply, what file structure exists, what metadata accompanies them, and what checksums were calculated before movement.

This creates a before-and-after comparison. If the transfer package arrives with different fixity values or a different file count, the archive knows the problem began during transfer rather than years later.

6. Ingest is the controlled entrance

Ingest is the process of taking digital material into managed archival custody. It commonly includes validating the transfer, identifying formats, checking for malware, calculating or verifying fixity, extracting metadata, recording provenance and placing the content into preservation storage.

Ingest creates a receipt. The archive should be able to answer: what arrived, when, from whom, in what condition, and what happened during processing?

7. Fixity answers: are the bits still the same?

A checksum is a calculated value derived from a file. If the file changes, the value usually changes. Fixity checking uses this property to detect unintended alteration or corruption.

The NARA Digital Preservation Program states that files should have recorded fixities to support auditing, preservation actions should be recorded and tracked, and regular audits should be performed. These are core archival controls because digital corruption can be invisible until a file is opened.

8. Fixity is not authenticity by itself

A checksum can show that a file has not changed since the checksum was calculated. It cannot prove that the file was authentic before that moment. If a false document enters an archive and receives a checksum, fixity can preserve the false document perfectly.

Authenticity depends on provenance, custody, recordkeeping process, metadata and documentation in addition to bit-level integrity.

9. Format identification is the first preservation diagnosis

File extensions are hints, not certainty. A file called report.pdf may not actually be a valid PDF. A digital archive uses format identification and validation tools to determine what the object really is and whether it conforms to known specifications.

Format knowledge matters because different formats carry different risks. Some are openly documented and widely supported. Others depend on one vendor, one software version or one operating system.

10. File formats have preservation risk

NARA’s Digital Preservation Framework assesses file-format risk and publishes preservation plans across record categories. The framework shows why “digital” is not one preservation problem. Email, images, software, databases, text, audiovisual files and web records have different significant properties and dependencies.

The correct preservation action depends on what must remain usable, not on a universal list of “safe” file extensions.

11. Significant properties define what must survive

If a file has to be migrated, the archive needs to know which characteristics matter. For a photograph, colour, resolution and dimensions may matter. For a spreadsheet, formulas and cell relationships may matter. For a database, table structures, constraints and relationships may matter. For a game or interactive artwork, behaviour may matter as much as visible content.

Preservation is therefore not always exact duplication of every technical characteristic. It is preservation of the properties necessary to sustain the record’s meaning, evidence and intended use.

12. Original files should usually be retained

When a preservation system creates a normalised or migrated copy, keeping the original gives future archivists a return path. A future tool may recover properties that today’s migration loses.

The Digital Preservation Coalition Handbook notes the importance of retaining the original format when migration is used, because transformations can introduce loss.

13. Preservation storage needs redundancy

One copy is not preservation. Multiple copies reduce the chance that a single disk failure, ransomware incident, building disaster, operator error or cloud-account failure destroys the archive.

Good redundancy also avoids a single failure domain. Three copies on the same storage array are not equivalent to copies separated across systems or locations.

14. Backup and archive have different jobs

A backup helps restore a recent operational state after failure. An archive preserves selected records over long periods with provenance and context.

Backups may overwrite old states, retain material only temporarily or optimise for fast recovery rather than permanent interpretability. Archives need deliberate retention, identity and preservation metadata.

15. Geographic separation reduces correlated failure

Fire, flood, power problems, local network compromise and physical theft can affect everything in one location. Preservation architecture therefore benefits from geographic separation appropriate to the institution’s risk model.

The Digital Preservation Coalition emphasises redundancy, media monitoring and checksums as foundations of bit preservation.

16. Storage media age

Hard drives fail. Tape ages. Optical discs delaminate. Solid-state media can lose charge. Controllers and connectors become obsolete. Cloud services change contracts and platforms.

Digital preservation therefore includes media refresh and storage migration. The archive must periodically move content to supported infrastructure before old media becomes unreadable.

17. Format migration protects renderability

Migration converts a record from an ageing or poorly supported format into a format more likely to remain usable.

Migration is powerful but not neutral. Formatting, embedded objects, formulas, fonts, scripts, colour profiles or interactive behaviour may change. Good preservation therefore tests migration against significant properties and documents the transformation.

18. Emulation protects behaviour

Emulation recreates an older computing environment so original files and software can behave in something close to their original context. It can be valuable for interactive works, software, games and complex digital objects where migration would destroy behaviour.

Emulation has its own dependencies: emulator software, operating-system images, documentation and rights. Preserving the object may require preserving a technical ecosystem.

19. Normalisation reduces future complexity

Some archives create preservation copies in a smaller set of well-understood formats. This reduces the number of formats the institution must monitor.

Normalisation should not erase the original. It is a risk-management layer: original received object plus preservation representation plus access derivative where appropriate.

20. Preservation master and access copy are different

A preservation master prioritises long-term quality and completeness. An access copy prioritises delivery, compatibility and reasonable file size.

The National Library Singapore’s digitisation specifications illustrate this distinction by using high-quality preservation formats and separate access derivatives for different material types.

21. Metadata is the memory of the preservation process

Digital preservation metadata can record identifiers, file formats, checksums, software, creation dates, relationships, rights, preservation events, agents and transformations.

Without this metadata, a future archivist may see two files and not know which is original, which was migrated, who performed the migration, what tool was used or what was lost.

22. PREMIS gives preservation metadata a vocabulary

The Library of Congress digital preservation programme identifies PREMIS as an international preservation-metadata standard designed to support long-term usability of digital objects.

The larger principle is more important than any one schema: the archive must preserve not only records but also the history of preservation actions performed on those records.

23. Provenance must survive automation

Automated pipelines can rename files, extract metadata, convert formats and generate derivatives at scale. Every automation step should leave a trace.

If the system changes ten thousand files overnight but cannot explain which tool ran, which settings were used, which outputs correspond to which inputs and which errors occurred, automation has increased opacity rather than preservation quality.

24. Databases are not just files

A database may depend on schema, relationships, stored procedures, code, application logic and external reference tables. Dumping the raw database file without documentation can make future interpretation impossible.

Preserving structured data therefore often requires data dictionaries, schema definitions, exports, documentation and evidence of how tables relate.

25. Email is a network of records

An email is more than visible message text. Headers, sender, recipient, timestamp, thread structure, attachments and mailbox relationships may matter.

Exporting email as screenshots preserves appearance but destroys much of its machine-readable structure. Archival preservation aims to retain the richest practicable representation.

26. Websites are moving targets

A website changes continuously. Pages call external scripts, databases, APIs, fonts, video platforms and advertising services. A capture made today may not render identically tomorrow.

Web archiving therefore captures bounded states over time. It may never preserve every interaction, but it can preserve enough structure and content to document a historical state.

27. Singapore’s web archive is part of legal-deposit continuity

Singapore expanded legal deposit to digital materials and web archiving in 2018. The National Library’s Web Archive Singapore preserves selected Singapore-related websites, while the legal framework allows archiving of websites containing the .sg domain within the statutory scope.

This is a useful example of how national memory systems adapt when publication moves from physical objects to changing networked environments.

28. Cloud-native records create dependency risk

Modern work increasingly lives inside platforms rather than files: collaborative documents, project-management tools, messaging systems, design platforms and cloud databases.

Archiving requires export planning before the service disappears. Does the platform export comments, version history, permissions and relationships—or only flattened content? A useful archive may need several export representations plus documentation of the original environment.

29. Encryption can preserve privacy and destroy access

Encryption protects sensitive records. It also creates a dependency on keys and key-management processes. If the archive preserves encrypted files but loses the keys, the bits survive but the records do not remain usable.

Long-term systems therefore need explicit policies for encryption, access control, key escrow, authorised decryption and future transfer of responsibility.

30. Privacy obligations do not disappear in the archive

Digital archives can make search and copying easier than physical archives. That increases privacy exposure. Personal data may need restriction, embargo, redaction or controlled access even when permanent preservation is justified.

Preservation and publication are separate decisions. A record may be preserved permanently while remaining unavailable to the public for a defined period or under defined conditions.

31. Cybersecurity is preservation

Ransomware, credential theft, malicious deletion and supply-chain attacks can destroy or corrupt digital holdings. Preservation therefore depends on security architecture: least privilege, monitoring, offline or immutable copies where appropriate, segmented systems, tested recovery and controlled administrative access.

The archive must defend both confidentiality and integrity. A system that prevents unauthorised viewing but allows silent mass alteration has not solved the preservation problem.

32. Disaster recovery must be tested

A backup is only a claim until restoration succeeds. Digital archives should test recovery procedures, document dependencies, know how long restoration takes and verify fixity after recovery.

NARA notes regular emergency backup restoration tests as part of its digital preservation activities. The principle is simple: resilience must be demonstrated, not assumed.

33. Access systems should not become the preservation system

Public websites and reading-room interfaces are designed for access. Preservation repositories are designed for controlled long-term management.

Keeping those functions distinct reduces risk. A public derivative can be replaced without altering the preservation master. A hacked website should not automatically compromise the authoritative archive.

34. Search can hide preserved material

Digital archives can preserve enormous quantities of material that remain hard to find. Search depends on metadata, indexing, OCR, entity extraction, access rights and interface design.

Preservation without discovery is incomplete for many public archives. Yet discovery tools must not replace archival structure. Users should be able to move from search hit back to collection, series, creator and record context.

35. AI can improve discovery and damage provenance

AI can transcribe handwriting, classify files, extract entities, summarise collections and help users navigate large record sets. It can also hallucinate, merge separate records, flatten uncertainty and detach an answer from its source.

An archive-aware AI should always preserve a return path to record-level evidence.

QUESTION
  → RETRIEVAL
  → RECORD IDS
  → PROVENANCE
  → RIGHTS CHECK
  → TRANSCRIPTION / OCR STATE
  → SYNTHESIS
  → CITATION
  → ORIGINAL OBJECT
  → USER CAN VERIFY

36. Preservation has an organisational layer

Digital preservation can fail even when technology works. Staff leave. Budgets disappear. Vendor contracts end. Documentation becomes stale. Governance becomes unclear.

A durable archive therefore needs designated responsibility, documented policy, succession, funding, risk review and a clear understanding of the community it promises to serve.

37. The designated community question

How much supporting information must be preserved depends partly on future users. Specialists may understand a scientific format with minimal explanation. The general public may need documentation, software or contextual metadata.

Preservation is therefore a promise to a future audience, not an abstract promise to bits.

38. Failure modes

Failure	What breaks
One storage copy	Single failure destroys the archive.
No fixity	Silent corruption can remain undetected.
Format unknown	Future rendering becomes unpredictable.
Metadata separated	Files survive without context.
Migration without logging	Transformations cannot be audited.
Preserve access copy only	Compression or convenience choices become permanent loss.
Cloud without export plan	Platform closure can strand records.
Encryption without key succession	Perfectly preserved ciphertext becomes unusable.
AI summary without record IDs	Answer loses evidentiary return path.
Backup never tested	Recovery may fail when disaster occurs.

39. A practical digital-preservation checklist

Know what records you are preserving.
Preserve provenance and transfer history.
Calculate and verify fixity.
Identify and validate formats.
Record technical, descriptive and preservation metadata.
Keep multiple copies across meaningful failure domains.
Retain originals when transformations are created.
Monitor format and media risks.
Document migrations and emulation dependencies.
Separate preservation masters from access copies.
Protect rights and privacy.
Test disaster recovery.
Keep the archive discoverable.
Plan organisational succession.

40. The deeper model: digital preservation is repeated proof

Physical archives often focus on slowing deterioration. Digital archives must repeatedly prove continuity:

STILL HERE?
STILL INTACT?
STILL IDENTIFIED?
STILL UNDERSTOOD?
STILL RENDERABLE?
STILL AUTHENTIC?
STILL ACCESSIBLE TO THE RIGHT PEOPLE?
STILL CONNECTED TO ITS CONTEXT?
STILL MANAGED BY A RESPONSIBLE INSTITUTION?

That repeated proof is the heart of digital preservation.

41. Why digital archives matter to civilisation

Modern societies increasingly conduct government, science, commerce, culture and private life through digital systems. If those systems leave no durable record, the historical archive of the twenty-first century can become paradoxically thinner than the archive of the twentieth.

Digital archives are therefore not a specialist technical concern. They are part of the infrastructure by which a civilisation remembers what it did.

To preserve a digital record is to preserve the possibility that a future person can still ask it a meaningful question.

Source and authority routes

Continue the Archives and Publishing series

Publication control: Wintour House · eduKate Publishing · evidence, edition, correction and archive gates.

World Return: If a digital record matters, do not ask only where the file is stored. Ask whether its bits, context, identity, rights, dependencies and preservation history will still make sense after the software, hardware and organisation around it have changed.

How Digital Archives Work | From Born-Digital Records to Fixity, Formats, Preservation and Future Access