Consider rendering web pages (XML, HTML, RDF, ...) in a 2D/3D vector space such represents each tag or pattern of tags by spatial elements, such as a lines, rectangles, cubes, spheres, etc... Any arbitrary representation (mapping data to virtual space objects) is valid, provided there is a way (additional widgets like buttons, sliders) for transforming an element into any other representation, or adjust its representation parameters, at any time.

Each spatial representation part can be "increased" or "decreased" – (ex: perhaps with small buttons attached, or dragging the border of the rectangle to resize, or simply clicking repeatedly). When increasing a certain section, it decreases everything else – in order that a quantifiable "attention" distribution is conserved among the tags of the page. The more attention that a tag receives, the larger, more vivid, or more "detailed" its spatial representation becomes, relative to the other tags that are depicted.

By the time one “leaves” a “site”, one will have sufficiently magnified the important parts of it. The resulting pattern, when combined with others', can be processed with a pattern-matching system that can be used determine default attention distributions for the pages one may visit. It may take into account what host it came from, the pattern of the URL, the time of day, or any other semantically meaningful variable.

Then, periodically combine, perhaps across a p2p network, a community of people's attention patterns to establish a "collaborative attention filter" that can be applied to each new page someone visits. As the filter becomes smarter and more sophisticated, the pages encountered will tend to magnify the content of the pages that previous visitors have decided "truly matters". The more agents participating in this system, the more demand will be placed on content providers to clean up their content, gradually minimizing "waste". People may even be employed by their communities to develop customized filters, further improving the process.

In addition to filtering or "weighting" page content, -- it may also generate annotations. This can employ existing semantic-web standards: each tag may be annotated by linking into to extra semantic content. Semantic content may be generated by rules or functions that create "semantic knowledge" automatically. Extra annotations and control widgets, if represented by objects in a zoomable space, avoid competition with content when “sized” relatively “smaller”.

Browse sentence by sentence or paragraph by paragraph. For example, maybe see one paragraph at a time, such that it consumes the entire screen. Then navigate through it with a tighter focus, at my own regulated pace, without being overwhelmed by oceans (pages and pages) of WORDS (which are discouraging to someone with ADD or dyslexia, which I suspect affects everyone to some degree or another) because it might cause anxiety of the possibility of ignoring something important, resulting in a loss of motivation to continue reading and exploring.

Also necessary is the ability to adjust the focus "window" or "aperture" size, so that any individual element or grouping of elements (words, sentences, paragraphs, images, etc) can be zoomed to consume the entire output display. The transition from one focus to another (when the focus's size changes, or its location changes) would be smooth and continuous in a zoomable spatial model. It is a model of "attention flux" – at any given time, it defines the size and shape of the "window" as well as all options for navigating to the subsequent focuses: whether to continue linearly, proceeding paragraph by paragraph, or to jump to a different section (as guided by a table-of-contents, or index) or a different site entirely (across a web hyperlink).

The resulting subconscious effect of displaying paragraph-level content to consume the entire screen (ie. Much bigger than "12pt-like" sizes employed by ordinary browsers and word-processors and e-mail software such as this) could be more impacting, resulting in an improved comprehension rate and quality.

By digesting the web in smaller chunks, as an interactive slideshow-like presentation in which each slide is not discretely defined but instead is an adjustable "focus window" that one can grow/shrink and move through multimedia objects in real-time. Determining what is visible or audible (ex: for the blind) or other modalities and qualities.

The fractal scale ranges from symbol/character extending to the entire page/document, and outwards to "sitemaps" (collections of inter-linked documents).