Glyphspace

Hwæt! For tonight I have been discussing e-ink, projected displays, and holographic movies with Pat, and as a consequence the future of computer interfaces. I'm enamoured with 2D interfaces, I make no secret of the fact; I think that even when 3D interfaces actually come along, for quite some time there'll be 2D interfaces that can blow them out of the water (hmm, it seems that epigram 10 has gone to my head). So with that in mind—and notwithstanding gauds, trinkets, and conceits like BumpTop which actually look very promising—I've been wondering about feasible 2D systems for the medium-term.

I suddenly had this daydream of an e-ink PDA, black and white, that organised data locationally. The idea of locationality and proximity is a recurring topic for me, because I like to diagram and move data about in 2D. I've often wondered whether an SVG program like Inkscape would be better for me to take notes in than emacs. The e-ink PDA I daydreamed ought probably be larger than a regular PDA, but not by too much. The main analogy of the interface is an OS map. For different scales there are different maps, showing only the most relevant data to that scale. If you had a vector map that was extemely detailed down to showing people's gardens etc., and then zoomed very far out, not only would it take a hideous long time to render but it would probably make the result look something like an aërial photo. Aërial photos are interesting in their own right, but aren't useful when, say, looking for motorways or major towns, airports, stations, etc. Indeed, even for a particular scale there are many possible useful representations, for example the common Political and Geographical with maps of the world.

So imagine if instead of the paradigm of solidus separated hierarchical names that denote discrete blocks of binary data (and, on Linux, other things such as pipes and devices) we had scribbles and a blank canvas the size of Prussia. You'd have to be able to prepare your own views of the data from 20,000ft, but you should be assisted in the task. There would be quite a reliance on handwriting analysis for data input, but I think that the original handwriting should be retained and the parsed text could annotate it, perhaps not being visible, or visible when you perform some action on it. Files could have their own little icons as usual, and be much like the GUI interfaces to files now.

But how would you run applications? I think that activities that you perform should be orthogonal to the locational aspect. Things are strewn across the landscape, but for example when you go to a tennis court you bring your own racquet and ball. I work a lot with logos and symbols, so that's a way that I'd like to invoke applications: by drawing symbols of what I want to do. So if I draw a magnifying glass, sort of like a backwards Q, that'd let me search. If I draw a little envelope, up comes my email application. A voice bubble? IRC. Clock? The current time. Musical notes? Audio application.

But how would these things even work? There could be entirely new approaches for them. On one scale, a little envelope in a box could signify an email, some kind of draft or a message received. But zoom in on it and it shows you the message text. Web servers could be implementing simply by sharing areas of the landscape. "Files" should be able to exist in many places at once, and there should be autocollations of sets of files. But more important than "files" are regions. You could draw paths that the PDA could then scroll you along, so that you could take a potted tour of various things you're working on; they'd be roughly analagous to roads. There could be points that you define where you could click to be taken to another place entirely, analagous more or less to airports. There could be subterranean layers exposing more of the backend of what's going on, the details—for the people who want to know how it all works. You should be able to dig under anything to find out what's going on, peeling off the surface to reveal diagrams representing the classes and objects of whatever language is powering the whole thing.

The history of the whole map could be backed up so that you could have an historical view of how your data has changed and moved about over time. No longer would you be deriving graphs and diagrams from your linear data, the graphs and diagrams would become first class data. You should be able to invent new symbols that mean particular things, and link them to certain behaviours. You should be able to use them as annotations as well as actions, though hopefully over time it will become apparent which kinds of actions are most useful and those could be built in. I'm not entirely sure how web browsing ought to work, but I wonder whether there could be a way to define a kind of meeting point or fountain where things group around, and then have a command to take you to a random bit of space close to this focus. Then you could draw a ruler indicating the display width, write in the URI you want to browse to, and draw something to indicate that it ought to be retrieved and rendered. You could give pages a temporal caching limit so that they disappear after N days, though with this system in general I'm assuming an almost unlimited amount of storage space.

Because typing is much quicker than handwriting, I think that the system should support being hooked up to a keyboard, perhaps one of those flexible portable keyboards, or even one of the projected ones though then you have to find a flat surface to work on. Otherwise I'm not really expecting too much in the way of frills. Given that I'm picturing this being an e-ink PDA, I'm not even sure that colour is all that important. This wouldn't be so multimedia inclined, and it's not a desktop computer replacement—rather it's the notebook to beat all notebooks before it. I once measured the information storage potential of a napkin and found that it could contain about 10kG, i.e. kiloglyphs; and the unit of measurement helped to crystalise my understanding that it's not just the depth of information storage that is important but the breadth too. There are many more possible glyphs than there are unicode characters. One should make a computer binarise as much of the input data as it can, e.g. with the handwriting recognition and so on, but not to constrain people to use only that.