Part 1 of 8 in The Text Document, a series on the history and philosophy of the textual document, from word processing to the cloud and machine authorship, and on the role open source and open standards have played in deciding who controls it.
In 1934, a Belgian lawyer named Paul Otlet published Traité de documentation, titled "the book on the book," aiming to rationalize the entire universe of recorded knowledge.1 A lawyer by training rather than a traditional librarian, Otlet had spent decades with Henri La Fontaine building the Mundaneum in Brussels. Together they assembled more than fifteen million index cards in wooden drawers, classified by the Universal Decimal Classification they co-invented, queryable by mail and telegraph decades before search engines ran on silicon.2 History would remember it as the "paper internet."
Otlet's ambition outran his century, and his institution nearly vanished in neglect and war. Yet the Traité left behind something more durable than cards in drawers: a working theory of the document. In 1937, working with international documentation bodies in Paris, Otlet settled a definition in three languages: a document is any source of information, in material form, capable of being used for reference or study or as proof.3
Three core requirements hide inside that sentence, and the ninety-year history of computer text is the story of their breaking and repair.
The antelope
The definition's sharpest elaboration came from Suzanne Briet, a librarian at the Bibliothèque nationale and Otlet's intellectual heir. In Qu'est-ce que la documentation? (1951), Briet mapped the boundaries of the concept.4 A star in the sky is not a document, though a photograph of it is. A pebble in a riverbed is not a document, though a stone catalogued in a mineralogical cabinet is. And then her famous example: an antelope. Running wild on the savanna, an antelope is just an animal. Captured, brought to Europe, placed in the Jardin des Plantes, catalogued, exhibited, and discussed in zoological monographs, that same antelope becomes a document. It has been made into physical evidence.
Briet's example is startling, but exact. The antelope enters documentation through its circumstances rather than its biology. It has been fixed in place, bounded, catalogued, and framed within an institution that makes it legible as proof of taxonomy and anatomy. Documenthood is not an inherent trait of matter; it is an evidentiary status imposed on it.
Information scientist Michael Buckland brought this Franco-Belgian lineage into modern computer science in his 1997 paper, "What is a document?", warning that digital technology "renews old questions and also old confusions between medium, message, and meaning."3 Buckland wrote at the height of the desktop office monopoly, when the world's working prose was disappearing into proprietary binary files. On paper, Otlet's definition had looked like library philosophy; inside a Windows 95 hard drive, it had become an engineering emergency.
From Otlet's definition and Briet's zoo, three load-bearing properties emerge. Every machine era of text either re-engineers them or trades them away.
The first is fixity: a document holds still. It has boundaries; it can be delimited, cited, archived, and compared across decades. A wild antelope cannot serve as proof because it runs away.
The second is evidence: a document testifies. It stands as witness to an author, a decision, a transaction, or an observation—what Buckland termed organized physical evidence.3
The third is independence: a document outlives the apparatus that made it. Once catalogued in the zoo, the antelope testifies to any visiting naturalist without requiring the original safari hunter in the room. A parchment charter does not need its dead scribe. Paper provided this third property so effortlessly that society mistook it for physics. Every digital format has had to fight for it.
When text moved to computers, all three broke.
The paper inheritance
Paper did so much architectural work for so long that its mechanics became invisible.
For millennia, "document" meant physical marks pressed into clay, papyrus, parchment, or rag paper. The three properties arrived pre-packaged in the physics of ink and fiber. A manuscript was fixed by sheets and binding. It testified through handwriting, ink formulas, wax seals, and watermarks. And it was independent because the sheet carried its own decoder: paper needed no runtime software, and a letter read identically in London and Lima.
Physical fixity was expensive, which was its greatest asset. Altering a contract meant visibly scraping vellum or striking through lines. Entire institutions, such as notaries, chancery clerks, and double-entry ledgers, arose simply to certify that an alteration had been authorized. Fixity and evidence were social contracts anchored in stubborn matter.
Paper's independence was never absolute. An illiterate peasant still depended on someone to read a Latin lease aloud. But the barrier was human literacy, not proprietary machinery, and the characters sat openly on the surface for anyone who knew the alphabet.
Computers severed that bundle in an afternoon: what a writer types, what a drive stores, and what a monitor renders became three separate layers. The industry has spent sixty years trying to reassemble them.
The opposite of a document
In 1960, as mainframes were first digesting text on punch cards, Ted Nelson began sketching the exact opposite of paper. Project Xanadu imagined writing not as bounded pages, but as an interconnected fabric: passages quoting other passages through permanent two-way links, every revision preserved, every excerpt dynamically pulled live from its source.5 Nelson coined hypertext for non-sequential writing, transclusion for real-time quotation without copying, and docuverse for the resulting literary commons.6
History often files Nelson as a frustrated prophet of the World Wide Web, which flatters the Web and misreads his design. Tim Berners-Lee shipped one-way links that break with 404 errors and abandoned transclusion and native versioning because simple files scaled and Xanadu did not. Nelson had articulated the anti-document: an architecture where page boundaries and frozen editions are obsolete constraints of printing, and writing remains an open, living stream.
The history that followed is a continuous clash between Otlet's instinct and Nelson's. The archival camp, including typesetters, John Warnock at Adobe, and the creators of PDF/A, worked to make bits hold still long enough to serve as reliable proof. The collaborative camp, including hypertext theorists, wiki designers, and cloud engineers, worked to dissolve pages into live, shared state.
In a modern cloud document flickering with remote cursors and automated drafts, neither vision has won outright.
The question sharpened
For most of my writing life, I rarely questioned the containers my words lived in. I drafted where it was convenient, assumed a saved file was an enduring possession, and trusted that an export menu would always bridge whatever gaps appeared. It took older files failing to open on newer computers, and collaborative cloud suites quietly converting working prose into rented application state, to reveal the deeper reality: the software we write with quietly decides who controls our record.
Buckland noted that moving from paper to bits revives ancient confusions between medium, message, and meaning.3 This series traces that tension through eight stages of computing history. The arc runs from IBM's $10,000 tape typewriter in 1964 and Xerox PARC's bitmapped screens to the binary .doc format, Adobe's frozen PDF, plain-text compilers like TeX and Markdown, standards battles over ODF, cloud workspaces, clean-room Rust rebuilds, and emerging disputes over machine-generated text.
At every stage, the test remains Otlet and Briet's: Does this text hold still? Can it stand as proof? And does it stand on its own, or only on loan from the vendor who built its container?
Briet's antelope entered documentation the day it was captured. This story begins in 1964, with an IBM machine built to capture something far more fleeting: a typist's keystrokes.
Next in Part 2, "The Text Gets a Body": from magnetic tape to Xerox PARC, tracing how WYSIWYG democratized writing while sealing text inside the application.
Notes and Sources
-
Paul Otlet, Traité de documentation: Le livre sur le livre, théorie et pratique (Brussels: Editiones Mundaneum, 1934; open digital edition, EMSHA, 2021). ↩
-
Mundaneum, "The Mundaneum: History"; Library of Congress, "Paul Otlet: Classifying Information, Counting Books, and the Internet" (May 2026). ↩
-
Michael K. Buckland, "What is a document?", Journal of the American Society for Information Science 48, no. 9 (1997): 804–809 (quoting the 1937 Paris International Institute for Intellectual Cooperation definition). ↩
-
Suzanne Briet, Qu'est-ce que la documentation? (Paris: Éditions Documentaires Industrielles et Techniques, 1951); English translation by Ronald E. Day, Laurent Martinet, and Hermina G. B. Anghelescu, What Is Documentation? (Lanham, MD: Scarecrow Press, 2006). ↩
-
Big Idea Initiative, "Xanadu: An Interconnected Fabric of Ideas". ↩
-
Theodor Holm Nelson, "Xanalogical Structure, Needed Now More Than Ever: Parallel Documents, Deep Links to Content, Deep Versioning, and Deep Re-Use", ACM Computing Surveys 31, no. 4es (1999). ↩
Top comments (0)