Ted Nelson named hypertext in 1965. His links were supposed to work both ways.
A manuscript-in-progress problem led to a design that treated revisions and reused passages as part of the reading experience.
You can follow a link to a page without that page keeping any record of the connection. Ted Nelson wanted the connection itself to be something a reader could inspect from either end.
His 1965 ACM paper, available in a reprint on Xanadu’s site, introduces hypertext alongside an information structure called zippered lists. Pieces of material could participate in different arrangements while retaining connections. An author could revise those arrangements as the work developed.
The paper begins with the writer’s predicament: notes become a manuscript without a neat dividing line. If you’ve ever tried to turn a folder of bookmarks into something coherent, that’s a fairly recognizable complaint to find in a paper from 1965.
Nelson’s proposed file structure included ways to preserve alternatives and changes. The thing you were writing could keep more of the decisions that produced it, much as a family recipe might keep the crossed-out ingredient that explains why this batch tastes different.
The paper distinguishes an information structure, a file structure and a proposed file language. Zippered lists belong to the first; the Evolutionary List File belongs to the second. That is more machinery than a blue underlined word needs, because Nelson was trying to manage the changing arrangements behind a piece of writing. A manuscript could have alternatives before its author had decided which version deserved to look final.
Nelson describes his earlier tools as physical things: card files, folders, scissors, paste and a roll-top desk. Those references make the imagined software easier to picture. Moving a passage shouldn’t have to erase the relationship it had with another passage, any more than putting an index card in a different pile should destroy the notes written on it.
The later model page proposes assembling documents from lists identifying existing content. Its account connects that reuse to the original addresses of the material, so excerpts can retain a route to their earlier context. The page also describes comparing connected documents side by side. This is the project’s explanation of its architecture, rather than a claim that ordinary hyperlinks cannot be supplemented by backlinks or revision history in other software.
That granularity changes what a useful citation might mean. A link to a whole page leaves the reader to find the passage and decide which version the writer saw. A connection tied to reused content could carry more of that relationship. The hard part for a new tool is keeping that promise as people revise, quote and rearrange the material after the first link has been made.
The project’s own assessment of the Web is blunt: “far too shallow.” That’s a later criticism from Xanadu’s model page. Treating the entire later publishing and copyright scheme as if it all appeared fully formed in the 1965 paper would flatten a long-running project into a much tidier origin story.
A new writing tool can make linking effortless and still leave the reader hunting for the original paragraph. Nelson’s older proposal keeps that question open: how much of a document’s history should arrive with the document?