A digital library has no requirement for a visible shelf.
That does not make it simpler.
The physical shelf once performed several jobs at once: storage, order, location and browsing. A digital library must rebuild all of those functions in software while also solving new problems such as file formats, authentication, broken links, rights management, integrity checking and long-term technological change.
This article is part of eduKateSG’s How X Works programme and the How a Library Works series.
The shortest useful answer
A digital library works by keeping digital objects under controlled identity, describing them with structured metadata, storing them in managed repositories, exposing them through search and browse systems, applying rights and access rules, and preserving both files and context so they remain usable over time.
The digital object is only one layer. The library also needs:
- an identifier;
- metadata;
- a storage location;
- a retrieval route;
- rights information;
- preservation information;
- an interface or machine endpoint;
- a record of relationships to other objects.
Without these layers, a folder full of files is not yet a strong digital library.
A digital file is not a library record
Imagine a server containing ten million files named with random strings.
The storage system may be functioning perfectly. Yet users cannot know what the files represent, who created them, what subjects they concern, whether two files are duplicates, whether a file is the latest version or whether access is permitted.
Digital librarianship begins when files become controlled objects inside a larger knowledge system.
Digital objects can be born digital or digitised
Digital libraries commonly contain two broad classes of material.
Born-digital material
This material originated in digital form: ebooks, websites, emails, datasets, digital photographs, word-processing files, software, audio recordings, videos or research outputs.
Digitised material
This begins as a physical object and is converted into digital form through scanning, photography, audio transfer or another capture process.
The distinction matters because a digitised surrogate is not identical to the physical original. Texture, scale, binding, annotations, paper, colour and other material evidence may not be fully represented.
The repository is the controlled storage core
A digital repository stores objects and the information required to manage them.
Unlike an ordinary shared drive, a repository can enforce identifiers, metadata requirements, version control, access rules, fixity checks, preservation actions and audit trails.
The repository is therefore the digital equivalent of stacks plus inventory plus part of the preservation room.
Ingest: the moment an object enters the digital library
Digital acquisition usually includes an ingest process.
- The object is received.
- Its identity and source are verified.
- File formats are identified.
- Viruses or technical hazards may be checked.
- Checksums are generated.
- Metadata is captured or created.
- Rights and access conditions are recorded.
- Files are stored in managed locations.
- Preservation copies or derivatives may be created.
- The object becomes discoverable according to policy.
Ingest is the digital equivalent of receiving, accessioning, processing and shelving combined into a software workflow.
Metadata is the object’s navigational envelope
Metadata allows the object to participate in search, management and preservation.
Different metadata layers may describe different aspects:
- Descriptive metadata: title, creator, subject, date and description.
- Administrative metadata: rights, ownership, access conditions and management details.
- Technical metadata: file format, dimensions, codecs, software dependencies or checksums.
- Structural metadata: relationships among pages, volumes, files or components.
- Preservation metadata: provenance, migrations, integrity checks and preservation actions.
This is why digital libraries depend heavily on metadata quality. The file can remain unchanged while discoverability, rights or meaning fail because the metadata layer deteriorates.
Identifiers separate identity from location
A raw web address often mixes identity with current location.
If the server changes, the URL can break even though the object still exists.
Persistent identifiers reduce this problem by giving the object a stable reference that can resolve to changing storage locations.
The deeper principle is important: a digital object should remain the same identified thing even when the infrastructure beneath it moves.
Search replaces only part of the shelf
Keyword search is powerful, but a good digital library offers more than a search box.
Users may need to browse by collection, subject, date, creator, geography, format or institution. They may need facets that narrow a large result set. They may need related-item links that recreate the intellectual neighbourhood once provided by physical shelving.
Digital browsing therefore depends on designed relationships rather than physical proximity.
Full-text search changes discovery
A printed book on a shelf can normally be searched only through its catalogue record until the user opens it.
A digitised book can potentially be searched word by word if its text has been transcribed or recognised through OCR.
This can transform historical research. A name buried on page 417 of an obscure report can become discoverable instantly.
But OCR is imperfect. Old fonts, handwriting, damaged pages and complex layouts can introduce errors. Full-text search therefore adds access while also adding another layer whose accuracy must be understood.
Derivatives: one object, several usable forms
A digital library may preserve a high-quality master file while delivering smaller derivatives to users.
For example, the preservation master might be a very large image file. The website may generate compressed JPEGs, thumbnails or zoom tiles for practical viewing.
This separates preservation quality from delivery efficiency.
The same object can therefore have several representations, each optimised for a different job.
Access control
Digital does not automatically mean open.
An object may be public, limited to registered members, restricted to a university community, embargoed until a future date, viewable only on-site, or inaccessible because donor or privacy restrictions apply.
Authentication establishes who the user is. Authorisation decides what that user may do.
These are separate questions, and digital libraries need both.
Licensing can make digital scarcity
Digital files can be copied cheaply, but licensed resources can still impose scarcity.
An ebook platform may permit only one simultaneous reader. A database licence may restrict users by institution. Streaming content may disappear when a subscription ends.
The technical capacity to duplicate does not override contractual rights.
This makes digital acquisition partly a rights-engineering problem.
Open access and institutional repositories
Universities and research institutions often operate repositories for theses, papers, datasets and other institutional outputs.
These systems preserve locally produced knowledge while making it discoverable beyond the institution.
The repository may expose metadata publicly even when the file itself is restricted.
This separation between description and access is important: knowing that an object exists is different from being permitted to open it.
Digitisation workflow
Digitisation is not simply pressing a scan button.
- Material is selected.
- Rights and preservation constraints are reviewed.
- The object is prepared and handled safely.
- Images, audio or video are captured to defined technical standards.
- Quality control checks completeness and fidelity.
- Files are named and structured.
- Metadata is created or linked.
- Masters and derivatives are generated.
- Objects enter repository storage.
- Discovery interfaces expose the collection.
Every skipped step can become tomorrow’s preservation or discovery problem.
Digital collections still need collection development
Storage feels cheap compared with shelving, but digital scale can become enormous.
High-resolution images, research datasets, video and web archives consume substantial storage and preservation resources. More importantly, every object creates metadata, review and maintenance obligations.
Collection development therefore remains essential. The library must still decide what deserves long-term institutional attention.
Digital preservation is continuous
A printed book stored well may remain readable for decades with little active intervention.
Digital objects depend on changing technical environments.
Digital preservation therefore uses redundant storage, checksums, format monitoring, migration, emulation, audit trails and preservation metadata.
The system repeatedly asks two questions:
- Are the bits still what we expect?
- Can current or future systems still interpret them?
Both are necessary.
Replication and redundancy
One digital copy is fragile even if the storage device is reliable.
Strong repositories maintain multiple copies across different systems or locations. This protects against hardware failure, local disasters, operator error and some cyber risks.
True resilience requires independence. Three copies on the same server are not equivalent to copies separated across failure domains.
APIs: when machines use the library
A digital library can serve humans through a website and machines through an API.
APIs can expose metadata, search results, images, identifiers or structured relationships so researchers and other systems can reuse the collection computationally.
This changes the library from a destination into infrastructure.
A user no longer has to visit one interface at a time. Software can query many collections, combine records and build new tools on top of them.
Interoperability
Digital libraries become much more powerful when their records can travel.
Shared standards for metadata, identifiers, rights and protocols allow aggregators to bring together collections from different institutions.
This is the digital equivalent of a library network: local collections remain separately governed but become traversable through common interfaces.
Linked data and knowledge graphs
Traditional catalogue records often package information inside one record.
Linked-data approaches represent people, works, places, organisations and concepts as distinct entities connected by typed relationships.
This can make digital libraries easier to traverse because the same person or concept can connect records across many collections.
The library begins to look less like a database of cards and more like a graph of cultural and intellectual relationships.
Authority control still matters online
Digital scale increases the cost of ambiguous identity.
If one author appears under five names, automated systems may split their work into five artificial people. If two people share a name, a digital library may merge them incorrectly.
Authority control gives the digital library a stable identity layer beneath changing names and labels.
User interface is only the surface
A polished digital library website can hide weak infrastructure.
If identifiers are unstable, metadata is inconsistent, files are missing, rights are unclear or preservation is neglected, a beautiful interface only makes the problems harder to see.
The strength of a digital library lies below the surface: in governance, metadata, storage, preservation and reliable routing.
Accessibility
Digital access can dramatically improve inclusion when designed well.
Screen-reader-compatible interfaces, transcripts, captions, alternative text, keyboard navigation, adjustable text, accessible document formats and downloadable data can open collections to users who face barriers in physical environments.
But inaccessible PDFs, image-only text and poorly structured interfaces can simply reproduce old barriers in digital form.
Privacy and user data
Digital systems can log searches, clicks, downloads, account details and reading activity at a scale physical libraries rarely could.
This makes privacy an architectural question.
The library should collect only the user data necessary to operate services, secure it appropriately and avoid turning reading behaviour into unnecessary surveillance.
Web archiving
Some of the most fragile modern publications are websites.
They can disappear when a domain expires, an organisation closes, a content-management system changes or pages are redesigned.
Web archiving captures selected sites over time so future users can study versions that no longer exist on the live web.
This is difficult because websites are not simple documents. They depend on scripts, databases, APIs, embedded media and external services.
Digital libraries and AI
AI can transform discovery by letting users ask natural-language questions across large collections.
It can generate synonyms, extract entities, summarise documents, recognise handwriting, classify images and connect related material.
But AI is most reliable when it operates over well-governed digital-library infrastructure.
Stable identifiers reduce entity confusion. Strong metadata improves retrieval. Provenance helps users distinguish primary material from generated explanation. Rights metadata tells systems what may be reused. Preservation records reveal whether the object has changed.
AI can become an extraordinary discovery layer, but it should not replace the evidence layer beneath it.
A digital library is a memory machine
At the deepest level, a digital library is a machine for preserving relationships through technological change.
The file may move servers. The interface may be redesigned. The storage vendor may change. The metadata schema may evolve. The user may arrive decades later through software that does not yet exist.
The library succeeds if identity, provenance, meaning and access survive those changes.
The complete mechanism
- The library defines what digital material belongs in scope.
- Objects are acquired, deposited, harvested or digitised.
- Ingest verifies files, provenance, rights and technical characteristics.
- Stable identifiers are assigned.
- Metadata describes identity, subject, structure, rights and preservation context.
- Repository storage maintains managed copies.
- Search, browse and APIs expose the collection.
- Authentication and authorisation control restricted access.
- Preservation systems check integrity and respond to format or infrastructure change.
- Interoperability connects the local repository to wider discovery networks.
- Usage and collection gaps feed back into future selection.
- AI and other interfaces can operate above the library while the evidence layer remains governed underneath.
That is how digital libraries work.
They do not abolish the old problems of librarianship. They translate them. Shelves become repositories. Call numbers become identifiers. Shelf browsing becomes graph traversal and facets. Conservation becomes fixity, migration and emulation. The medium changes, but the mission remains: keep knowledge identifiable, reachable and usable across people and time.