A location tells you where something is now. A persistent identifier is designed to keep telling you what the thing is after “where” changes.
Web URLs are powerful because they combine identity and location. They are also fragile when sites migrate, publishers reorganise, storage paths change or resources move between systems.
Persistent identifiers separate the enduring identity of a resource from its current network location. The identifier remains stable; a resolver or registry updates the location and metadata behind it.
This article sits beneath How Metadata Works, How Citation Works and How Entity Resolution Works. Metadata describes resources. Citation points to them. Entity resolution reconciles identities. Persistent identifiers provide a governed identity anchor that can survive change in location.
1. Identity and Location Are Different Problems
A URL such as example.org/files/report-2026.pdf tells a network client where to request a resource. If the file moves, the address can stop working.
A persistent identifier says, in effect, “this is the identity of the resource; ask the resolution system where the current representation lives.”
The distinction is the same one seen throughout this series: identity should not be allowed to drift merely because one implementation detail changed.
2. Persistence Is a Governance Promise, Not a Property of the String
No identifier becomes persistent by looking permanent.
A DOI, Handle, ARK or other persistent identifier works only because organisations maintain registries, resolution infrastructure and policies for updating destinations.
The string can remain unchanged for decades only if someone keeps the mapping current.
3. Resolution Connects Identifier to Current Location
A resolution service receives the identifier and returns or redirects toward the current resource location or metadata record.
If the resource moves, the registry changes the target rather than minting a new identity for the same object.
4. Metadata Makes the Identifier Useful Beyond Redirects
A persistent identifier can be associated with title, creator, date, publisher, version, type and other metadata.
This lets discovery systems recognise what the identifier refers to even when the full object is unavailable or moves again.
5. Versioning Must Be Explicit
Does a new edition keep the same identifier or receive a new one?
The answer depends on the identifier policy and the scholarly or operational meaning of version identity. A corrected file can be the same intellectual object in one system and a distinct citable version in another.
Persistence therefore requires rules for when change preserves identity and when change creates a new entity.
6. Granularity Determines What the Identifier Owns
An identifier can refer to a whole dataset, one release, one record, one image, one chapter or one article.
Too coarse and citation cannot point to the exact object. Too fine and the system creates unmanageable identifier proliferation.
The right granularity follows the unit that needs independent identity, citation and lifecycle management.
7. Persistence Does Not Guarantee Availability
An identifier can continue resolving to a metadata record after the object itself becomes restricted, withdrawn or unavailable.
Persistence means the identity remains traceable. It does not guarantee perpetual public access to the content.
8. Tombstone Pages Preserve the Record of Absence
When a resource must be removed, a persistent identifier should not simply disappear if the citation record matters.
A tombstone page can explain that the object was withdrawn, why where appropriate, and which metadata or replacement resource remains available.
This preserves scholarly and institutional memory instead of turning withdrawal into unexplained broken linkage.
9. Persistent Identifiers Support Citation Networks
When cited objects have stable identifiers, reference systems can connect publications more reliably across publishers and databases.
That supports citation tracking, provenance and evidence graphs. The canonical citation owner remains How Citation Works.
10. ORCID-Like Identity Shows the Same Principle for People
Researchers can change institution, email address and name format. A persistent person identifier lets systems keep one scholarly identity across those changing attributes.
This is entity resolution strengthened by a governed identity token.
11. Persistent Identifiers Reduce Link Rot — but Only With Maintenance
Ordinary links break when paths move and redirects are not maintained. Persistent identifiers reduce that risk by making location change an expected maintenance event.
If the registry itself is abandoned, persistence ends. The organisational layer remains the final dependency.
12. Worked Example: Research Article
A journal article receives a DOI. The publisher later redesigns its website and moves the article to a new path.
The DOI does not need to change. The registry target is updated so old citations can still resolve to the current article location.
13. Worked Example: Dataset Versions
A research dataset is updated each year.
The project may maintain one concept-level identifier for the dataset family and separate identifiers for individual releases. That allows a reader to cite exactly the state used in an analysis while still navigating to the continuing project.
14. Worked Example: Museum Object
A museum object can move storage location, gallery, conservation lab or lending institution while retaining one accession identity.
The accession number behaves like a local persistent identifier: location changes, identity remains.
The museum estate already owns the collection-specific lifecycle through articles such as How Museum Works | Accessioning.
15. Worked Example: eduKateSG Canonical Routes
A knowledge estate benefits from stable canonical URLs because internal links, external references and search systems accumulate around them.
If a route must move, redirects preserve continuity. A formal persistent-identifier layer would take that idea further by separating content identity from any one site path.
16. Persistent Identity and Canonical URL Are Related but Different
A canonical URL says which accessible web address should represent a resource in the current web estate. A persistent identifier can sit above location and resolve to whichever URL currently hosts the resource.
The previous How URL Canonicalization Works article owns the web canonicalisation mechanism.
17. Failure Mode: Identifier Changes With Every Migration
If the identifier is regenerated whenever the storage location changes, it is acting as an address rather than persistent identity.
The whole point is to absorb infrastructure change underneath a stable reference layer.
18. A Persistent-Identifier Checklist
- Define the entity or resource that deserves stable identity.
- Separate identity from current location.
- Choose the appropriate granularity.
- Define when a changed version keeps or receives a new identifier.
- Maintain metadata beside the identifier.
- Operate or use a durable resolution service.
- Update targets when resources move.
- Preserve tombstone records when objects are withdrawn.
- Monitor broken resolution and stale metadata.
- Treat persistence as an organisational commitment, not a syntax feature.
19. Read the Mechanism Forward, Backward and Sideways
Forward: resource → persistent identifier → registry metadata → resolver → current location. Backward: start from a broken citation and ask whether the location changed without the identity mapping being maintained. Sideways: compare author, archivist, publisher, database and reader. Each depends on the same stable identity for a different downstream job.
20. The Civilisation Lesson
Civilisation depends on references surviving change. Books move libraries. institutions rename themselves. websites migrate. storage media disappear. people change employers.
Persistent identifiers are one way to preserve continuity through that movement: the infrastructure can change while the record still knows what thing was being referred to.
A persistent identifier is a promise that the identity of a resource will outlive the place where the resource happens to live today.
Return through How Knowledge Graphs Work, How Metadata Works, How Citation Works and the master How X Works hub. Together, the Language & Representation corridor follows grounding, notation, invariant preservation, multimodality, annotation, identity, controlled terminology, semantic compression, knowledge graphs and persistent reference.