A library looks quiet from the outside. Inside, it is moving all the time.
Books arrive. Journals are licensed. Websites are assessed. Records are created. Subjects are assigned. Shelves are rearranged. Files are migrated. Loans begin and end. Questions arrive at a desk, through a search box, or through an API. Materials are repaired, withdrawn, digitised, preserved, replaced and sometimes rediscovered after decades of neglect.
The useful question is therefore not simply what is a library? It is how does a library keep knowledge usable across people, places and time?
This article is part of eduKateSG’s How X Works programme. It treats the library as a mechanism: a governed system that takes a very large world of possible information and turns part of it into an organised, retrievable, trustworthy and reusable collection.
The shortest useful answer
A library works by repeatedly solving nine connected problems:
- What should enter the collection?
- How should it be acquired?
- How should it be described?
- Where should it sit in relation to everything else?
- How should people discover it?
- Who may access it, and under what conditions?
- How should it survive physical or digital decay?
- How should the institution know whether the collection is still useful?
- What should happen when an item leaves, changes, or returns in another form?
These are not separate housekeeping chores. They form a loop. Acquisition without cataloguing creates invisible material. Cataloguing without classification creates description without structure. Classification without discovery creates order that users cannot reach. Discovery without preservation creates access that expires. Preservation without selection can conserve enormous quantities of material that nobody has reason to keep.
A library begins before anything is on the shelf
The first act of a library is exclusion.
No library can contain everything. Even a national library, a web archive, or a large research institution must decide what is in scope. A neighbourhood public library, a primary school library, a university medical library and an engineering company’s technical archive are all libraries, but they make different selection decisions because they serve different missions.
This is why a collection is never neutral in the mechanical sense. It is shaped by budget, audience, legal obligations, language, curriculum, research priorities, community need, preservation capacity, licensing rules and institutional purpose.
The collection-development question is therefore not “is this item good?” but “does this item deserve scarce attention, money, storage, description and future maintenance here?”
That decision leads directly into library acquisition: the mechanism through which desired material becomes controlled institutional material.
Acquisition turns possibility into custody
Libraries acquire material in many ways: purchase, subscription, donation, legal deposit, exchange, licensing, digitisation, transfer, deposit agreements and direct creation of institutional records.
Acquisition is not merely paying for a book. It is the transition from we want this to we are responsible for this.
For a printed book, that responsibility may include receiving, checking, accessioning, ownership marking, processing and shelving. For an electronic journal, it may involve a licence, authentication rules, user limits, coverage dates, renewal dates and perpetual-access clauses. For a digital archive, it may require format checks, fixity verification, metadata capture and storage redundancy.
Custody is therefore broader than possession. A library may own an item but fail to make it usable. It may provide access to a resource it does not own. It may preserve a digital object that is legally restricted. It may hold a rare object that can only be consulted under supervision.
Cataloguing gives an item an identity
Once material enters the system, the library must say what it is.
This sounds easy until two books share a title, one work appears in twenty editions, an author writes under several names, a journal changes title, a conference proceeding is issued by an organisation that also changes name, or a digital object has no obvious physical container.
Library catalogues solve identity and relationship problems. A catalogue record can describe a work’s title, creator, edition, publication details, physical or digital form, subjects, identifiers, notes, access conditions and relationships to other works.
Good cataloguing reduces ambiguity. It helps the system distinguish “the same work in another edition” from “a different work with a similar title.” It helps users move from an author to all associated works, from a subject to related material, and from one manifestation of a work to another.
In effect, cataloguing converts an object into a node in a knowledge network.
Classification gives the collection shape
Cataloguing asks, “What is this?” Classification asks, “Where does this belong?”
In a physical library, classification may determine shelf order. In a digital library, the same principle may appear as subject hierarchies, taxonomies, facets, controlled vocabularies, tags, ontologies or graph relationships.
A useful classification system compresses complexity. It gives users a way to move from the general to the specific, and from one item to nearby items. It also creates intellectual neighbourhoods: books about thermodynamics sit near related physics; materials about constitutional law sit near legal structures; a user looking for one concept can encounter another that is meaningfully adjacent.
This mechanism is explored in depth in How Library Classification Works.
The shelf is a map, not just storage
A shelf is easy to underestimate because it is physical. But shelf order is a user interface.
When classification, call numbers, signage and shelving are coherent, the shelf supports two forms of discovery at once. The first is precise retrieval: “take me to this item.” The second is neighbourhood discovery: “show me what sits around this idea.”
Digital systems reproduce the same logic with search results, related-item panels, subject facets, citation links, recommendation systems and browse trees. The medium changes. The routing problem does not.
Discovery is the visible surface of invisible work
Most users meet a library through a search box. That search box is only the surface.
Behind it are decisions about metadata, indexing, spelling, language, relevance ranking, authority control, synonyms, subject headings, access rights, database coverage and record quality.
A search for “heart attack” may need to reach material catalogued under “myocardial infarction.” A search for a person may need to recognise initials, former names, transliterations and variant spellings. A search for a historical country may need to bridge older geographic terminology. A search for “climate change” may need to retrieve material published before that phrase became standard.
This is why discovery quality depends on work done long before the search begins.
Circulation is a controlled movement system
When a library lends material, it temporarily separates an item from its normal location while trying to preserve its future availability.
That requires identity, state and time. The system needs to know which copy is being borrowed, by whom, when the loan began, when it is due, whether it can be renewed, whether another user has reserved it, whether it is overdue and what happens when it returns.
A circulation system is therefore a state machine. An item may move through states such as available, on loan, in transit, reserved, missing, under repair, reference-only or withdrawn. Policies define which transitions are allowed.
The same principle governs electronic access. Instead of a barcode, the system may authenticate a user account, institution, network address or licence entitlement. Instead of a due date, the restriction may be simultaneous-user limits, download rights or subscription coverage.
Reference work solves the problem the catalogue cannot see
Users rarely arrive with perfectly formed questions.
A student says, “I need something about pollution.” A researcher asks for “the report Singapore used before that policy changed.” A parent remembers a children’s book by its cover colour. A historian has an incomplete citation. A doctor needs evidence quickly but cannot afford to search every database separately.
The reference function converts an uncertain human need into a tractable search route. Librarians clarify the question, identify likely source types, choose databases, refine terminology, test results and help the user judge evidence.
This is one of the deepest reasons libraries remain useful even when information is abundant. Abundance does not remove the routing problem. It increases it.
Preservation fights several kinds of decay
Libraries exist partly because memory is fragile.
Paper can become acidic, bindings can fail, ink can fade, insects can damage collections, water can destroy storage, and repeated use can wear out a book. Digital material faces a different but equally serious set of threats: obsolete file formats, failed disks, corrupted bits, lost passwords, expired domains, discontinued software, broken links and vanished vendors.
Preservation therefore means preserving usability, not merely preserving an object. A digital file that still exists but can no longer be opened is only partially preserved. A rare book locked away without an adequate record may survive physically yet remain functionally invisible.
Good preservation strategies include suitable environmental control, handling rules, conservation treatment, digitisation, redundant storage, integrity checks, format migration, documentation and disaster planning.
Libraries are governed by permissions and obligations
Access is never just technical.
Copyright, privacy, donor restrictions, national law, institutional policy, licence agreements, archival sensitivity and community expectations all shape what a library can do with material.
A library must therefore balance openness with stewardship. The ideal of access matters, but so do legal rights, confidentiality, cultural sensitivity and the long-term survival of fragile material.
This makes the library a governance system as much as an information system.
Physical libraries and digital libraries solve the same deep problem
A physical library has shelves, rooms, desks, trolleys and return chutes. A digital library has databases, indexes, servers, authentication layers, metadata stores, resolvers and interfaces.
At first they look different. Mechanically, they are close relatives.
- The shelf has a digital equivalent in organised storage and browse structures.
- The catalogue card becomes a metadata record.
- The call number becomes a persistent identifier or structured locator.
- The circulation desk becomes an access-control and entitlement system.
- The reserve shelf becomes a digital hold, queue or restricted-access state.
- The preservation room becomes redundant storage, checksums, migration and backup.
- The reference librarian becomes guided search, research support and increasingly AI-assisted routing.
The important distinction is not physical versus digital. It is whether the system can maintain identity, context, discoverability, authority and access through change.
Why libraries classify people’s questions as well as materials
A mature library learns that users do not search only for objects. They search for outcomes.
“I need a book on fractions” may actually mean “I need to help my child understand equivalent fractions tonight.” “I need articles about urban heat” may mean “I need enough evidence to design a research question.” “I need old newspapers” may mean “I am trying to verify whether a family story is true.”
The visible request is often an imperfect proxy for the real task. This is why good library systems combine object-level organisation with user-level interpretation.
The library as a network of relationships
A modern library is less like a warehouse and more like a graph.
An author connects to works. Works connect to editions. Editions connect to publishers. Topics connect to broader and narrower topics. Articles connect through citations. Archival records connect to creators, organisations, places and events. Items connect to collections. Collections connect to institutions. Institutions connect through interlibrary loan, shared catalogues and consortia.
Once these relationships are represented, the library can support more than lookup. It can support traversal.
A user can move from a known book to its references, from a scientist to collaborators, from an event to primary sources, from a subject to neighbouring subjects, or from a local record to a larger international catalogue.
This is the same deeper idea explored in How a Mechanism Library Works: knowledge becomes more powerful when it is not merely stored but made traversable.
What makes a library trustworthy?
A library does not guarantee that every item is true. Libraries contain disagreement, obsolete science, propaganda, fiction, error, historical prejudice and competing interpretations.
The trustworthiness of a library comes from a different place: provenance, description, context, repeatable retrieval, transparent organisation and responsible stewardship.
A trustworthy library helps a user know what an item is, where it came from, how it relates to other material, whether it has been altered, what access conditions apply and how to locate it again.
Trust is therefore infrastructural.
Libraries also forget
Collection growth cannot continue without limit. Libraries withdraw duplicate, damaged, superseded, unused or out-of-scope material. They cancel subscriptions. They replace editions. They deaccession items under policy. They move low-use material to storage. They sometimes retain one copy in a preservation repository while removing many local copies.
This is not necessarily failure. Forgetting can be a form of maintenance when it is governed carefully.
The dangerous form of forgetting is uncontrolled loss: a file disappears because nobody renewed a domain; a local publication is never collected; a catalogue record is deleted without preserving its history; a rare item decays because its significance was never recognised.
The feedback loop: use changes the library
A library is not complete when material is acquired. It must learn from use.
Loan data may reveal demand. Search logs may reveal failed queries. Reference questions may expose gaps. Course reading lists may create pressure for more copies. New research fields may require new journals. Community change may create new language needs. Damaged books reveal handling and replacement requirements. Digital-access statistics can reveal heavily used resources and expensive subscriptions with little value.
These signals feed back into selection, budgeting, metadata improvement, service design and preservation.
The collection is therefore not a static stock. It is a continuously governed model of what an institution believes should remain reachable.
Libraries cooperate because no library is enough
One of the most powerful ideas in librarianship is that a local collection does not need to contain everything if it can reliably reach other collections.
Union catalogues, shared metadata, interlibrary loan, consortia, reciprocal access, digital repositories and standards allow libraries to behave as parts of a larger network.
This changes the meaning of collection strength. A strong library is not merely one that owns a great deal. It is one that knows what it owns, knows what it does not own, and has routes to obtain what its users need.
What changes when AI enters the library?
Artificial intelligence changes the discovery layer more quickly than it changes the underlying obligations.
An AI system can help users translate vague questions into search terms, compare sources, generate query variants, identify entities, cluster documents and traverse relationships. But it still depends on the quality of the library beneath it.
If metadata is weak, provenance is missing, rights are unclear, identifiers are unstable or collections are poorly connected, AI does not remove the problem. It can hide it.
The strongest future library therefore combines machine-scale discovery with library-grade stewardship. AI can accelerate routing; the library must still preserve identity, evidence and trust.
The complete mechanism
We can now read the whole machine in one pass.
- A community or institution defines a mission.
- Selection decides what deserves collection attention.
- Acquisition brings material under institutional control.
- Cataloguing establishes identity and descriptive context.
- Classification places the item in a larger intellectual structure.
- Storage and preservation keep the material viable.
- Discovery systems expose it to users.
- Circulation and access rules control movement and use.
- Reference services translate uncertain questions into workable routes.
- Usage, gaps and failures feed back into future collection decisions.
- Cooperation connects the local library to a larger information world.
- Governance keeps the whole process lawful, accountable and durable.
That is how a library works.
It does not merely keep books. It keeps relationships between knowledge and people alive long enough for one generation to find what another generation discovered, wrote, measured, imagined or learned.