The canon

Global name for any resource, from a web page to a person or an idea; the foundation the rest of the stack builds on. IRIs extend URIs to Unicode.

Long-running W3C Technical Architecture Group issue over whether an HTTP URI names a web page or the thing the page describes, such as a person.

Tim Berners-Lee's 1998 note arguing that a URI, once published, should keep working indefinitely, and that breaking links is a choice made by poor design rather than an accident of technology.

Roy Fielding's June 2005 message announcing the W3C TAG's answer to httpRange-14: a URI that answers with a 2xx response names an information resource, while one that answers with a 303 redirect may name anything.

W3C Interest Group Note (2008) giving publishers two recipes for URIs that name things rather than pages: hash URIs, and 303 redirects based on the httpRange-14 resolution.

2010 ISWC paper showing that owl:sameAs links between Linked Data sets are often used to mean 'similar' or 'related' rather than strict identity, and proposing finer-grained alternatives.

The US grocery industry's product barcode and numbering standard, chosen on 3 April 1973 from IBM's design; the first scan was a pack of Wrigley's gum in Troy, Ohio, on 26 June 1974.

Persistent identifier for documents and data, introduced in 2000. A DOI names the object rather than where it is stored, and resolves through doi.org; it is built on the Handle System and fits within the URI system.

W3C Technical Architecture Group's account of the Web's architecture: identifying resources by URI, interacting with them through representations, and data formats; a W3C Recommendation since 15 December 2004.

GS1 standard for writing product identifiers such as GTINs as web URIs, so that a barcode or QR code on a pack resolves to information about the product; ratified in August 2018. It puts Linked Data identifiers on everyday goods.

W3C specification for identifiers that their subject can create and control without a central registry. DID 1.0 became a Recommendation in July 2022; DID 1.1 reached Candidate Recommendation in March 2026.

Database of films, television and the people who make them. It began in October 1990 as fan lists on the rec.arts.movies newsgroup, became a company in 1996 and was bought by Amazon in 1998; its data is proprietary, released only for non-commercial use, yet other databases link to its title and name IDs.

Database of more than 25 million place names, founded in 2005 and published under a Creative Commons attribution licence. Each place has a stable URI that returns a web page or RDF, with links to Wikipedia, which made it the usual place identifier in linked data.

Free, collaboratively edited knowledge graph run by the Wikimedia Foundation. It supplies structured data to Wikipedia and the sameAs links on this site.

Essay in The Atlantic Monthly imagining machines that extend human memory, including the memex, a desk for storing records and linking them by association.

Douglas Engelbart's 1962 SRI report setting out a research programme to use computers to extend human problem-solving, which led to NLS.

Ted Nelson's paper at the 1965 ACM national conference, which introduced the word 'hypertext' in print.

Tim Berners-Lee's proposal to CERN management for a linked information system. It became the World Wide Web.

Roy Fielding's 2000 doctoral dissertation at UC Irvine, which defined REST, the architectural style of the Web: resources named by URIs, manipulated through representations over a uniform interface. Most Web APIs since claim descent from it.

Tim Berners-Lee's short Design Issues note setting out four rules for publishing data on the Web with URIs and links.

Survey by Bizer, Heath and Berners-Lee of the principles and early growth of the Web of Linked Data.

2009 ESWC paper by Georgi Kobilarov, Tom Scott, Yves Raimond and colleagues at the BBC and Freie Universität Berlin, describing how BBC Programmes and BBC Music used DBpedia and MusicBrainz identifiers to connect content across the BBC's site.

HyperText Markup Language, the markup of Web pages, whose anchor element carries the Web's links. Berners-Lee first described it in the informal 'HTML Tags' document in late 1991; HTML 2.0 was published as RFC 1866 in November 1995.

Hypertext Transfer Protocol: the Web's request–response protocol for fetching a resource by its URL. Berners-Lee's first version was written up in 1991; HTTP/1.0 and 1.1 followed as IETF RFCs. Looking up a URI to get data about the thing it names depends on it.

Shared web APIs for digitised images, proposed in late 2011 by research libraries including the British Library, Stanford, the Bodleian and the BnF. Its Image API (2012) serves zoomable images, and its Presentation API (2013) describes books and artworks as JSON-LD manifests, so any viewer can show any library's images.

Ted Nelson's hypertext project, begun in 1960, aiming at a 'docuverse' with two-way links and transclusion. Only partial versions have been released.

Douglas Engelbart's oN-Line System at SRI, with hypertext links, a mouse and shared screens; publicly shown on 9 December 1968.

Hypertext system of documents linked by URLs and served over HTTP, built by Tim Berners-Lee and Robert Cailliau at CERN from 1990.

Project that extracts structured data from Wikipedia infoboxes and publishes it as Linked Data; an early hub of the Linked Open Data cloud.

E.F. Codd's 1970 paper in Communications of the ACM proposing that data be organized as relations (tables), independently of how it is stored.

Marvin Minsky's MIT AI Lab memo introducing frames: structures of slots with default values for stereotyped situations.

Peter Chen's paper in the first issue of ACM Transactions on Database Systems, proposing to model data as entities and the relationships between them.

1997 submission to the W3C by R.V. Guha and Tim Bray describing MCF in XML; a direct precursor to RDF.

Google blog post by Amit Singhal announcing the Knowledge Graph, which brought the term into wide use.

Survey by Aidan Hogan and 17 co-authors of the data models, query languages, reasoning and methods behind knowledge graphs.

Functional Requirements for Bibliographic Records: IFLA's conceptual model of the bibliographic universe as entities and relationships, separating the work from its expressions, manifestations and items; its final report was published in 1998.

Ontology for museum and cultural heritage documentation from ICOM's documentation committee, first issued in 1999 and an ISO standard since 2006. It describes objects through the events in their history, such as production, acquisition and movement, so that records from different institutions can be integrated.

Data model that describes resources as subject–predicate–object statements identified by URIs. It became a W3C Recommendation in 1999 and was revised as RDF 1.1 in 2014.

Collaboratively edited knowledge graph built by Metaweb and bought by Google in 2010. It fed Google's Knowledge Graph, and its data was offered to Wikidata when it closed in 2016.

MAchine-Readable Cataloging: formats for library catalogue records, developed by Henriette Avram at the Library of Congress between 1965 and 1968 so that libraries could share records by computer; a US national standard by 1971 and international by 1973. MARC 21 (1999) is still the predominant format.

Guidelines for encoding texts for scholarly study, from a 1987 meeting of humanities and computing scholars at Vassar College; P3 appeared in 1994, P4 moved to XML in 2002 and P5 followed in 2007. It is the standard markup of the digital humanities.

W3C markup language for structured documents and data, a Recommendation since February 1998; RDF's first official syntax was written in it.

Family of XML formats for syndicating site updates. Netscape's RSS 0.90 (1999) and RSS 1.0 were RDF-based; RSS 2.0 is plain XML.

Conventions for marking up people, events, reviews and similar data with plain HTML class names, without a separate data format.

JSON-based syntax for Linked Data that lets ordinary JSON be read as RDF; a W3C Recommendation since January 2014.

John McCarthy's 1959 paper proposing the 'advice taker', a program that would decide what to do by deducing it from facts and rules stated in formal logic; the founding statement of logic-based knowledge representation.

1970 paper by Waldo Tobler in Economic Geography, a computer animation of Detroit's growth, in which he states the first law of geography.

William Woods's 1975 critique showing that the links in semantic networks had no agreed meaning, mixing up, for example, a class with its members and a definition with an assertion; it set the agenda for giving networks a formal semantics.

Brachman and Schmolze's 1985 account of KL-ONE, a frame-based system with defined concepts and an automatic classifier that places each concept in a subsumption hierarchy; KL-ONE was the first knowledge representation system based on what became description logics.

Levesque and Brachman's 1987 paper showing that reasoning gets dramatically harder as a representation language gets more expressive, and that this trade-off underlies the differences between existing representation formalisms.

Clay Shirky's essay arguing that the Semantic Web's reliance on syllogistic logic fails on real-world, context-dependent data.

Design history of OWL by Horrocks, Patel-Schneider and van Harmelen, tracing it from description logics and RDF.

Halevy, Norvig and Pereira's 2009 essay arguing that simple models trained on very large amounts of web data beat elaborate hand-built models, and that meaning is better learned from use at scale than specified in advance.

2012 NIPS paper by Alex Krizhevsky, Ilya Sutskever and Geoffrey Hinton whose deep network, trained on GPUs, won the ImageNet image-recognition contest by a wide margin and turned machine learning toward deep learning.

2013 NIPS paper introducing TransE, which represents each entity in a knowledge graph as a vector and each relation as a translation between vectors, so that missing facts can be predicted; the starting point of knowledge-graph embeddings.

2013 paper by Tomas Mikolov, Kai Chen, Greg Corrado and Jeffrey Dean introducing word2vec: vectors learned from which words occur together, in which relations such as king to queen show up as directions.

2017 paper by Ashish Vaswani and seven Google colleagues introducing the transformer, a network built on attention that became the basis of large language models.

W3C language for writing ontologies on the Web, based on description logics. OWL 2 followed in 2009.

Long-running project to hand-encode common-sense knowledge as a logical knowledge base, started at MCC and continued by Cycorp.

Raymond Reiter's 1978 paper formalizing the closed-world assumption: a database may treat any fact it cannot show to be true as false.

Paper by Pérez, Arenas and Gutiérrez giving SPARQL a compositional formal semantics and proving the complexity of evaluating it; general SPARQL patterns are PSPACE-complete.

Query language for RDF data, a W3C Recommendation since January 2008; SPARQL 1.1 followed in 2013.

OGC standard, version 1.0 from 2012 and 1.1 from January 2024, that adds a vocabulary for geometries to RDF and spatial functions and relations to SPARQL, so queries can ask what is within, near or touching what.

W3C language for validating RDF data against shapes, closer to schema checking than to OWL's open-world reasoning; a Recommendation since July 2017.

2004 OSDI paper by Jeffrey Dean and Sanjay Ghemawat describing how Google split computations over huge datasets across thousands of ordinary machines; it set the pattern for big data processing.

ISWC 2013 paper that monitored hundreds of public SPARQL endpoints and found many of them often unavailable, slow, or limited in what they would answer: evidence that free public query services were hard to keep running.

2016 Journal of Web Semantics paper proposing Triple Pattern Fragments: servers answer only simple triple patterns and clients do the rest of the query, trading some speed for servers that are cheap to run and stay up.

Paul Otlet's treatise on organising documents and knowledge, setting out the field he called documentation.

Tom Gruber's paper on sharing ontologies between knowledge systems. It contains the definition of an ontology as 'an explicit specification of a conceptualization'.

Clay Shirky's 2005 essay, from talks at O'Reilly ETech and IMCExpo, arguing that hierarchical classification is a habit of the shelf and that on the Web links and free tagging work better than expert ontologies; the best-known case for folksonomy.

2007 paper by Yves Raimond, Samer Abdallah, Mark Sandler and Frédérick Giasson at ISMIR presenting the Music Ontology, a framework for editorial, cultural and acoustic information about music on the Semantic Web.

2008 EKAW paper by Martin Hepp presenting GoodRelations, an ontology for describing what businesses offer, at what price and on what terms.

Library classification that sorts knowledge into ten main classes with decimal subdivisions, first published by Melvil Dewey in 1876.

S.R. Ranganathan's library classification, which builds subject numbers from facets joined by colons.

Small set of metadata elements, such as title, creator and date, for describing resources on the Web; it began at a 1995 workshop in Dublin, Ohio.

RDF vocabulary for describing people, their online accounts and who they know; one of the first widely used Semantic Web vocabularies.

W3C vocabulary for describing classes and properties of RDF data; the first schema language for RDF.

W3C vocabulary for publishing thesauri, classification schemes and subject headings as Linked Data; a Recommendation since August 2009.

Shared vocabulary for structured data in web pages, launched in 2011 by Bing, Google and Yahoo, with Yandex joining later that year.

Lexical database of English begun in 1985 at Princeton's Cognitive Science Laboratory under George A. Miller: nouns, verbs and adjectives grouped into sets of synonyms linked by relations such as broader and part-of. It became a shared vocabulary for language processing and for knowledge bases such as YAGO.

Ontology of gene and gene-product functions begun in 1998 by researchers working on fly, mouse and yeast genomes; one of the most used ontologies in biology.

Paul Baran's 1964 RAND memorandum comparing centralized, decentralized and distributed networks, and showing that a distributed network survives attacks on its nodes and links far better than a centralized one.

Juan Benet's 2014 paper describing the design of IPFS.

Brewster Kahle's talk to the NetGain gathering at the Ford Foundation. He argues that the Web is fragile and lets readers be tracked and censored, that 'the way we code the Web will determine the way we live online', and calls for a peer-to-peer Web with no central host.

Ink & Switch essay setting out seven ideals for software that keeps data on users' devices while still allowing real-time collaboration.

W3C Recommendation (2018) for decentralised social networking, used by Mastodon and the wider Fediverse. Its messages are Activity Streams 2.0 in JSON-LD, so the Fediverse runs on Linked Data.

Tim Berners-Lee's project for storing personal data in user-controlled 'pods' that apps read and write through Linked Data standards.

Brin and Page's 1998 paper describing Google and PageRank, which ranks pages by the links pointing to them; it showed that the Web's link structure could serve as a signal of authority, with no metadata needed from publishers.

WWW 2005 paper extending RDF so that a set of statements can be named by a URI and described like any resource. It shows publishers signing their graphs and consumers deciding which graphs to trust.

2010 paper by Paul Groth, Andrew Gibson and Jan Velterop in Information Services & Use defining the nanopublication.

Platform for Internet Content Selection: W3C labels describing the content of web pages, made to let parents filter what children see; the label specification became a Recommendation on 31 October 1996. The RDF specification calls it a predecessor of RDF.

W3C family of specifications for recording provenance: who made something, from what, and how. PROV-DM is the model, PROV-O its OWL ontology; both 2013.

W3C data model for tamper-evident digital credentials, such as diplomas or licences, that can be checked cryptographically. Version 1.0 became a Recommendation in November 2019, and version 2.0, with its Data Integrity and cryptosuite specifications, in May 2025.

Email and file encryption program released by Phil Zimmermann in 1991. Instead of a central certificate authority, users vouch for each other's keys, the model its 1992 manual named the web of trust.

Linked Data published under an open licence. Berners-Lee's 2010 five-star scheme rates open data from 'available in any format' up to 'linked to other data'.

Principles that research data should be findable, accessible, interoperable and reusable, published in Scientific Data in March 2016.

1999 book by Lawrence Lessig on how the Internet is regulated, arguing that software architecture ('code') constrains behaviour much as law, markets and norms do.

Principles written by thirty open-government advocates meeting in Sebastopol, California, on 7–8 December 2007: government data should be complete, primary, timely, accessible, machine-processable, non-discriminatory, non-proprietary and licence-free.

2008 Communications of the ACM article by Weitzner, Abelson, Berners-Lee, Feigenbaum, Hendler and Sussman arguing that the Web should make the use of information accountable to rules, rather than rely on hiding and access control alone.

2009 book by Viktor Mayer-Schönberger arguing that digital memory has made forgetting the exception, and proposing expiration dates for stored information.

Judgment of the EU Court of Justice of 13 May 2014 (C-131/12) holding that search engines must, on request, remove links to information about a person that is inadequate, irrelevant or no longer relevant.

Open Knowledge Foundation's definition of 'open' for content and data: anyone can freely access, use, modify and share it, subject at most to requirements that preserve provenance and openness. First drafted in 2005.

Open, volunteer-edited database of music metadata, launched in 2000 in reaction to CDDB's enclosure. Its core data has been in the public domain since 2003, and its identifiers anchor the Music Ontology, BBC Music and Wikidata's music items.

Free, collaboratively written online encyclopedia launched on 15 January 2001 and run by the Wikimedia Foundation. Its infoboxes and links are the raw material of DBpedia, YAGO and Wikidata.

Volunteer-made map of the world, started in 2004 because Britain's Ordnance Survey would not release its data under a free licence. Its features carry free-form key=value tags, a folksonomy steered by a wiki, and its data moved to the share-alike ODbL in 2012.

Cory Doctorow's essay arguing that reliable metadata is unattainable because people lie, are lazy and disagree.

2019 Communications of the ACM article by engineers from Microsoft, Google, Facebook, eBay and IBM comparing how their companies build and run knowledge graphs, and what remains hard.

Diagram of datasets published as Linked Data and the links between them, first drawn in 2007 with DBpedia at its centre.

Google search feature announced in May 2009 that shows ratings, prices and other details in results, read from microformats and RDFa in pages.

1978 paper by Donald Kessler and Burton Cour-Palais in the Journal of Geophysical Research predicting that collisions in low Earth orbit would come within decades and that debris would overtake meteoroids as the main hazard by about 2000.

Book edited by Charlotte Hess and Elinor Ostrom applying Ostrom's research on how communities sustain shared resources to knowledge: digital libraries, open access, open-source software and the rules that keep them from being depleted or enclosed.

2014 PLoS ONE study of millions of web links in scholarly articles, finding that about one in five articles cites web resources that have since vanished or changed: link rot measured at scale.

Nadia Eghbal's 2016 report for the Ford Foundation arguing that open-source software is public infrastructure, maintained by a few unpaid or underpaid people, and that nobody has taken responsibility for funding it.

WWW 2016 paper on moving Freebase's data into Wikidata after Google announced in December 2014 that it would close Freebase: what could be carried over, what could not, and why a company's knowledge base did not simply transfer to a community one.

Persistent Uniform Resource Locator: a web address that points at a resolver instead of the resource, so the target can move without breaking links. OCLC introduced it in 1995 and ran purl.org, where Dublin Core and many other vocabularies keep their namespaces, until it handed the service to the Internet Archive in September 2016.

The Internet Archive's public archive of the Web, open since 25 October 2001 and built from crawls going back to 1996; for many dead links and vanished sites it is the only copy left, kept by a nonprofit funded mainly by donations and grants.

Manu Sporny's January 2014 blog post, written days after JSON-LD became a W3C Recommendation, saying that better Web APIs, not the Semantic Web, motivated JSON-LD, and that keeping RDF out of sight was what made it usable.

Apple's programmable card-and-stack application for the Macintosh, written by Bill Atkinson. It put linked, scriptable documents in the hands of ordinary users.