home > notes > hypertext
v1.0 11aug2001: First version
By Matt Webb, who in the bath this morning was having a think and realised that his ponderings about hypertext appear to have satisfactorily concluded, in that a decent metaphor, or collection of metaphors, have been arrived at, after about a month or more of reading, discussing and thinking about text, the mind, and UI.
1. No text is complete ^
There is no text that can be read without reference to other texts. A piece of text is a foam of metaphors and implications. Reading is filling this in.
In difficult to understand texts, the references must be more explicit (footnotes), but if these are not enough the reader will look for context. In the situation of a page, references may be made by physical proximity.
A very incomplete text is one replete with references.
2. There is a whole graduation of links ^
Links between texts run along a scale of implicit to explicit. Normal texts contain implicit links (or references, or metaphors); the register of the voice of the writing is an implicit link. Www links are the crudest form of explicit link. Other links made be made to paragraphs, entire documents, or as citations. The placing of a text (in physical/www space, or in time as a presentation) is an implicit link, or context.
It could be argued that a single tract of text is a mush of ideas, implicitly linked by proximity.
Social conventions are massively important in referencing.
3. The node/arc model is insufficient ^
The node/arc model says that a piece of text is a node, and a reference is a link. Links have no further attributes other than connecting two pieces of text. (The node/arc model also has a place in UI, or in people interactions, or in any kind of network -- it's a very powerful metaphor). The node/arc model says that a node is given extra definition by the nodes connected to it.
This model only works for links that are forks or branches -- abrupt and explicit links. A text which builds a complex idea and implicitly references another text with this idea is more a slow forking of the flow. Texts can be closely links, or juxtaposed. Links are given context by the text they're in and have meaning.
A different model could follow the reader who would explore textual links, following idea branches, as if flowing down a river.
The www is the strongest adherent to the node/arc model, but the appearance of the web to work like this is reinforced by the people building it using this model internally. Really a single page could be seen as many many nodes and arcs, tightly bound, with links to other pages being very long arcs. That there aren't medium length arcs is an accident of www link design.
The node/arc model also does not account for connections of more than one nodes; paths through the hypertext hopping from node to node. There are no supernode structures in this model.
4. In the future, linear text will be seen as a special case ^
All text is part of the universal hypertext. A linear text, like a book, is simply a text with no explicit references, and all the forks chosen by the author. Really, a linear text is a trail.
5. The most adequate model is the most general combination, because hypertext is real ^
There is no single model for hypertext, only a series of ways of describing its different aspects. We [humankind] store our ideas in some kind of hyperspace. We don't store them in the same way; we don't share the same hypertext; our own hypertexts are probably internally inconsistent.
But because there is in fact this thing I call hypertext, this idea store, these words and links and metaphors, it's real. And because it's real, it's irreducible and indescribable.
Any complex system has emergent properties. It's impossible to describe any system outside its own terms, because outside those properties don't exist (try explaining envy from the perspective of an electron). Hypertext has its own terms which make it impossible to talk about in a complete, yet simplified, way.
6. Trails are how we read ^
Trails are paths through hypertext. Trails follow the flow, slowly forking at links. Trails are a structure in hyperspace, that space in which all texts exist. They can be as simple as a list of other texts, or they can be long tracts, referencing many other texts all the way.
Each trail is a dimension in the hypertext, a dimension that has direction (or order). ZigZag from Project Xanadu is almost like this, more so that the web. It connects nodes in long trails (dimensions). Xanadu is an implementation of an idea which acknowledges trails.
7. Trails are a form of text ^
All trails are texts, and all texts are trails. Although texts can be seen as nodes, they could also be seen as a path or journey through all the other texts in hyperspace (this word I'm using for the space in which the entire structure of the universal hypertext exists). If a trail is a line, a text is a way of looking at this line end on, so it appears as a point.
A book is a form of trail, following implicit links, following the textual flow.
Trails needed be linear: they can be complex investigative structures, playing the flow around many different texts.
8. Human interaction is like a flow hypertext ^
People aren't links, and interactions aren't arcs. Proximity is important; previous links are important; internal structure is important. Human interaction is just an explicit links in amongst all the internal implicit links. Human interaction and thought is the building up of this massive internal hypertext, where parts only come into existence as they're explored.
9. A shared hypertext is just trails ^
A hypertext which is shared and external, like the entire body of world literature, or the www (they can't really be differentiated as separate hypertexts, but it's interesting to contrast their different structures), doesn't pre-exist to being authored.
Everybody lays a trail through the shared hyperspace as they author or read. Reading is a way of crossing somebody else's trail. We call crossing points nodes, and we call explicit steps on the path arcs.
On the www: Consider every browser history there every way, each one a chronological trail. Some of these steps are authoring, some reading. A point of reading is a webpage. When a link is followed, that's an explicit step (an implicit step is looking up a page in a search engine, or guessing a url). Consider the mesh of these trails in hyperspace, then flatten it so you only have the knots and the explicit steps left. And that's the web.
home > notes > hypertext