A history of hypertext before the Web

This text about the early history of concepts of ‘hypertext’ and its first implementations before the World Wide Web was launched, was originally prepared in October 2016, to share with the members of the KIDMM email discussion list.

Part of this account is from memory, part is paraphrased from the book ‘Multimedia and Hypertext’ by Jakob Nielsen, and Wikipedia helped to fill in the gaps.

Vannevar Bush and the Memex (1939–45)

Nielsen considers that the basic ideas of hypertext, though not its name, were developed by Vannevar Bush, who wrote about a conceptual machine called the Memex (memory extender). These ideas were developed in the early 1930s, first emerged as a draft paper in 1939, and were finally aired in public in an article in Atlantic Monthly called ‘As We May Think’ (1945).

Vannevar Bush was a scientist, indeed a government scientific adviser, already concerned about the explosion in scientific information and journals. How could scholars and scientists keep on top of this?

The Memex was conceptualised as a mechanised private file and library, storing papers and correspondence and other previously paper-based documents on microfilm. Frames from these films could, he imagined, be projected on screens for reading, with several such reading windows, so comparisons and links could be made. It would be possible to buy publications on compatible microfim format.

However, Bush envisaged that the user would also be able to add handwritten marginal notes and comments and other materials, via some unspecified imaging technology. And the core feature of the Memex was to be what he called ‘associative indexing’. As he explained: ‘The basic idea... is a provision whereby any item may be caused at will to select immediately and automatically another. This is the essential feature of the memex. The process of tying two items together is the important thing.’

Ten Nelson and Xanadu (1965)

The next step along was taken by Ted Nelson, and Jakob Nielsen dates this to 1965, which is when Nelson gave a talk about his ideas to the Association for Computing Machinery, and invented the term ‘hypertext’ (Nelson invented a lot of words…) At this time, his ideas included a computer-mediated way of showing text, and allowing it to be edited – this was before word processing was invented – but also allowing different versions of it to be compared.

He was also conceiving the idea of access to a universe of texts, which he later called the ‘docuverse’, implying a system of local and remote server-stored text databases; the ability to put together compound documents from any selection of subdocument text strings (he called these ‘zippered lists’ and the process of inserting a text fragment into a compound document ‘transclusion’), and the possibility of nonsequential document construction.

In 1967, Nelson gave his idea a name: Project Xanadu. The ideas got a public presentation in two books: the 1974 compound book ‘Computer Lib / Dream Machines’ and the 1981 ‘Literary Machines’. In the 1970 and 1980s there were also attempts to realise Xanadu, or at least a demonstration thereof, in software, and for a while the project was financed by Autodesk.

Xanadu included some powerful ideas which I don’t believe were ever successfully implemented. One of these was that in linking to a text, you could link not just to the document, but to any range of text within it. This seems to imply either being able to write markers into the target text, or more likely some sort of out-of-line link database. Another idea was that links which had been established between texts could be followed from either end. A third idea was some sort of automatic linkage to all the versions of a document that had ever been, including the links that had been established to each version, allowing a kind of temporal scrolling.

Nelson and his followers seem to have been miffed at being ‘pipped to the post’ by Tim Berners-Lee and HTTP/HTML, launching a tirade of scorn against such a ‘stupid’ system with one-way links to dead pages.

Hypertext Editing System (1967)

Nelson was associated with Brown University in Providence, Rhode Island, and that institution has other links to hypertext systems. Jakob Nielsen claims that the world’s first working hypertext system was created at Brown in 1967 by Andries van Dam: the Hypertext Editing System. This research was funded by IBM and ran on a memory partition of a small IBM/360 mainframe. It was actually sold by IBM to the Houston Manned Spacecraft Center, and was used to create documentation for the Apollo programme.

A follow-up hypertext-like project at Brown was FRESS, the File Retrieval and Editing System.

HyperTIES

Ben Schneiderman, who is perhaps known to some as the author of ‘Designing the User Interface’ and the ‘treemap’ system for visualising volumes of hierarchically structured information, also developed a hypertext system, called HyperTIES, at the University of Maryland in 1983. The earliest versions ran in DOS text mode, without a scrolling display but rather a series of fixed pages linked in a book. Hyperlinks were followed by moving the cursor to a link text and pressing the Enter key. A later version ran on Sun workstations using the NeWS window system.

HyperTIES was commercialised by the Cognetics Corporation. It had a small feature set but did have touchcreen support, and was best used for information display kiosk applications. Many of the information systems authored for HyperTIES presentation were produced by museum curators and intended for use by visitors to the museum.

One unusual feature of HyperTIES was that following a link did not necessarily take you straight to the destination: a hypertext collection could be authored so that activating a link would first bring up a short synopsis or definition at the bottom of the screen, offering the option to switch to a full ‘article’ on that topic. As for the jump link, it took you to the first page of that article, not to a specific page, meaning that you might have to page through several screens to find what you were after.

NoteCards (1984?)

The University of Maryland was where Randall Trigg wrote the first PhD thesis about hypertext, in 1983, and then he went off to Xerox’s famous Palo Alto Research Center (PARC) and worked with a couple of other guys to create a new hypertext system called NoteCards.

NoteCards had four kinds of objects: the cards themselves (presented as windows on the computer screen), links between them, a special ‘browser card’, and a FileBox object to manage a collection of cards and their links. FileBoxes could contain other FileBoxes, recursively.

NoteCards was written in Lisp and ran on Xerox InterLisp D-machines with a high resolution B/W bitmapped display and a mouse, the sort of thing which Apple Macintosh was at that time bringing to the desktop (but on the Mac, with a much smaller screen).

The ‘cards’ themselves, as I said, were windows, contained within the larger FileBox window; the cards could contain text and graphics, and they could be resized, but did not have a scrolling behaviour. You could have many open at once, a notable contrast with HyperTIES, and they could overlap and be selectively brought to the front, but you had to beware of a ‘messy desktop’ developing. Links between cards were signified by a small icon embedded in the text, and clicking on a link icon opened another card as a new card window.

The Browser Card contained a system-generated boxes-and-connectors diagram representing all the cards in a FileBox and the links between them. It could be used to leap to a particular card, or to operate on the link structures.

…and the Instructional Development Environment

The InterLisp environment meant that users who were skilled in Lisp programming could extensively customise the functions and appearance of NoteCards, and one team at PARC thereby developed IDE, the Instructional Design Environment, specifically aimed at helping courseware authors to construct hypertext systems semi-automatically. An interesting innovation in IDE was the ‘Fat Link’, a link icon pointing to several different cards. Clicking on a Fat Link brought up a menu of available links.

Intermedia (1985)

Meanwhile in 1985, back on Rhode Island and at Brown University, Norman Meyrowitz and colleagues developed another hypertext system for developing courseware: Intermedia. It was used at Brown to teach several courses in the humanities and the natural sciences. It ran under A/UX, Apple’s first foray into Unix, and was quite resource intensive for its day, needing something like a Mac II with 4 MB of RAM and 80 MB of disk space (I know, I know, but that was a lot back then).

Because of its courseware paradigm, Intermedia was built on the assumption that an authored set of hypertext documents would be shared by a group of students, who should be able to create their own links in addition to the centrally authored ones, and should also be able to annotate the documents for their own purposes. Therefore, the architecture of Intermedia stored separate files for each student, holding that student’s own links and annotation cards.

The linking process in Intermedia drew on the ‘copy-paste’ Mac analogy, so that one would start a link from one object such as a user-defined text or graphic ‘Block’, navigate to the link destination, select or make another Block, and then use the Complete Link command to establish the link. As with NoteCard, the existence of a link was advertised with a small icon. As for appearance, Intermedia benefitted from the Mac System-level font resources. It was quite a promising system, but the project ended in 1991 when funding was withdrawn.

Symbolics Document Examiner (1985)

Also in 1985, Lisp workstation manufacturer Symbolics funded the development by Janet Walker of the Symbolics Document Examiner. Symbolics had a very specific purpose: to convert their 8,000-page manual for their sixth release of their Genera object-oriented operating system into an on-screen hypertext reference of 10,000 ‘hyperdocument’ nodes lashed together with 23,000 links.

This specific purpose gave the Document Examiner some peculiar properties. Firstly, the book metaphor mean that it had a hierarchical structure – something other hypertext systems have specifically tried to get away from. Authoring had to be done in another environment (Symbolics Concordia). All the user could do was to add bookmarks to make it easier to return to specific pages. Nevertheless the system was well liked by its users, and won an award from the Society for Technical Documentation.

Texinfo

(As a ‘footnote’, it seems that Janet Walker was in part inspired by the Texinfo system developed as an offshoot of EMACS by Richard Stallman. This, I believe, is an open-source version of LaTeX for documentation, the current version of which either emits a DVI file for printing, or a hypertext manual generated in HTML form.)

OWL Guide (1982–87)

Since about 1982, Peter Brown at the University of Kent had been developing a hypertext system called ‘Guide’. Initially this ran on Three Rivers PERQ workstations running Unix. Office Workstations Ltd (OWL) became interested in it as a commercial product, and in 1986 they ported Guide to Apple Macintosh, and in 1987 to Microsoft Windows, making it the first hypertext system to run on both platforms.

Guide was both an authoring and viewing/using environment. Quite different than most hypertext systems, Guide conceived of the document as essentially linear, but with sections which were hidden from view until a link was invoked. That operation caused the document to ‘break apart’, inserting the additional material as a kind of transclusion – not that dissimilar to Ted Nelson’s concept of StretchText, I suppose. Closing the link ‘vanished’ the additional material. Thus writing for Guide made one think in terms of overviews and details. However, Guide also allowed for ‘jumps’, which would take the user off to another document altogether; and for very small annotation, Guide also supported ‘pop-ups’. The form of link was indicated by the mouse cursor changing its appearance.

Over time, OWL moved Guide in the direction of authoring multimedia, but I don’t know much about that. Guide version 2 introduced a fourth kind of link that executed a script written in ‘Genesis’ – usually to access a clip from a videodisc.

BBC Domesday Project

Also about this time, we should mention the BBC Domesday Project, which engaged schoolchildren and others across Britain in providing text and images and videos, which were converted into a kind of hypertextual delivery system written in BCPL (Basic Combined Programming Language, a forerunner of C). The data was stored on Laservision Read Only Memory disks, like CDs but the size of a long-playing gramophone record, and the player was controlled via a SCSI card by a BBC Model B computer.

HyperCard (1987)

And that brings us to 1987 and HyperCard on the Apple Macintosh. Bill Atkinson, who had created MacPaint for Apple, started work on this in 1985, apparently after an inspiring experience with lysergic acid diethylamide. Atkinson came to an agreement with Apple that he would give HyperCard to Apple only if they would give it away for free.

It turned out to be wildly popular, and all sorts of people used it to create or prototype applications, build databases, and all kinds of things. Apple had an ambivalent attitude to HyperCard: on the one hand it helped make Macs popular, on the other, they earned nothing from it and it undermined the market for various sorts of shrink-wrapped software produced by Apple’s developer allies. Apple finally killed off HyperCard in 2000.

Jakob Nielsen argues that HyperCard was not strictly a hypertext system, because the links and interactive elements were not attached to text per se, but to regions on the ‘card’. The metaphor was of a stack of fixed-size cards, similar to index/filing cards. Each card had a layered structure, with the rearmost plane inheriting characteristics such as shared visuals and I think navigation from a template card. On top of that could be placed text boxes and graphic objects, which might provide the visual form of buttons, and then on top there were live regions, clicking on which would activate the link to another card in the stack, or to some other desired function.

Two powerful features of HyperCard got people up to speed quickly. First was the provided library of well-designed graphical user interface objects, and second was the scripting language, HyperTalk. HyperTalk is a rather verbose language (Wikipedia gives the example put the first word of the third line of field "hello" into field "goodbye", but this makes it easier for non-programmer mortals to figure out how to use it. HyperTalk was also an interpreted language, until version 2.0 brought an on-the-fly compiler.

One person whose thinking about hypertext was stimulatedby his experiences with HyperCard was Robert Cailliau, who is not often remembered as Tim Berners-Lee’s collaborator on the World Wide Web. Likewise, Brendan Eich, the developer of JavaScript, was inspired by HyperTalk. Ward Cunningham, who invented wikis, also refers to a stack he wrote in the late 1980s. So HyperCard has been pretty influential.

I’ve certainly had some experience with HyperCard. I also used to have three of Don Norman’s books in HyperCard form, released by the Voyager company.

Apple never devoted much programmer resource to HyperCard and this, plus its extinction in 2000, its limitation to the Macintosh platform, and failure to become compatible with Mac OSX, made room for competition and for successors. Most of these I know *of* but not *about* – e.g. SuperCard, Plus, and LiveCode.

  • SuperCard launched with native colour support and its scripting language is a superset of HyperTalk; I believe that it is still going well as version 4.7.3, and runs under Mac OSX.
  • LiveCode was developed by Runtime Revolution (in Edinburgh), and was formerly called Revolution and then MetaCard. Very similar in concept to HyperCard, and inspired by it, LiveCode has a scripting language called Transcript, and runs on a very wide range of platforms – Windows, Mac, Linux, Solaris, Raspberry Pi, iOS and Android – even the Web. It is often used to develop mobile applications.

FrameMaker hypertext

I’ll mention two other hypertext systems of which I’m aware, and then stop. Version 3 of FrameMaker, when it was still a Frame Technologies product, introduced hypertext navigation features, including pop-up lists of links, and some companies used it to develop on-screen manuals. There is a reader application for FrameMaker, called FrameReader, which is still available from Adobe for Windows and Macintosh, and which can use these hypertext features. A subset of these features can be exported to make interactive PDF.

DynaText

Finally, I’ll return us to Providence, Rhode Island, where that developer of the 1960s Hypertext Editing System, Andries van Dam, teamed up with Louis Reynolds, Steven DeRose and Jeffrey Vogel to form the company Electronic Book Technologies. The software they developed, DynaText, converted SGML documents into a binary representation plus a full-text inverted index of the text, elements and attributes. The inspiration was arguably similar to that of the Symbolics Document Examiner. Customers typically distributed DynaText production of CD-ROM, and I believe it was used for such things as on-line documentation of weapons systems and aircraft.

I was certainly impressed by DynaText when I saw it being exhibited by Lou Reynolds at a conference in Amsterdam of the Graphic Communication Association, a conference about the Standard Generalized Markup Language (SGML) which I attended, I think in 1991.

Also at that conference I met Sharon Adler and Anders Berglund: I was in a workshop session they jointly led, about the Document Style Semantics and Specification Language, DSSSL. They were both members of the ISO committee which had ushered SGML into the world as ISO 8879; and lest anyone think that ISO Committees are unromantic, I must report that Anders and Sharon later got married.

…and so to the World Wide Web

But I bring Anders into this story because at the time he was the Publishing Manager for a European physics laboratory called CERN, and it was he who championed the use of SGML in CERN’s publishing operation. And I didn’t know it, but back in Geneva, two of Anders’ colleagues, Tim Berners-Lee and Robert Cailliau, had just knocked up their own hypertext system, the World Wide Web, based in part on an SGML-inspired markup language. But that story is surely well enough known for me not to have to tell it.

    — Conrad Taylor, October 2016