Files and Text

As Ted Nelson points out, the files are currently chunks of data owned by particular applications and the operating system has no access to their contents. This could be superseded by a text-oriented approach that would make the operating system responsible for managing a text structure including all software and data available. Imagine a text-engine that would store a text structure as presented in this book. Imagine that the entered text could be efficiently queried and navigated. The file system would be replaced by the text structure itself: the operating system would know not only about unit names, but also about roles and types. Instead of running ls /usr/bin one would use ls :program to list all units of type ”program“ or

ls =job.#document."Jane" modified

to get all documents under the name ”job“ that were modified by Jane. But one would also get the first paragraph of each article with ls :article.(1)p and all references to Xanadu inside them with

ls :article.#p."Xanadu" ref

An application such as a word processor would not own a chunk of data any more. To present a particular book the user is working on, the application would retrieve it from the operating system by getting

=my-new-book.:prose

that would return a list of all prose elements. The application would render it and let the user update it. But the same book could be operated by say a spreadsheet application that would be interested in

=my-new-book.:chapter.:status

that is in the titles of the book's chapters and their status: ”work in progress,“ ”review“ etc. and would show these in tabular form for the user to manage the project of writing the book. In this scenario there are still separate applications with particular purposes, but the information they work on is unique and perfectly integrated.