The Form Factor
The fastest way of entering text is still the keyboard. People have tried to convince me that handwriting recognition is getting good, but most humans have problems with my handwriting, and it gets worse every year. Speech recognition is now pretty good for a single speaker, after some training, but dictation is really hard. There is a reason few executives have Dictaphones anymore: It's faster for them to type their own letters than to dictate the letters and have a secretary type them. This wasn't the case with mechanical typewriters, where a single mistake could mean rewriting the whole page, but after the invention of the Backspace key, the situation quickly changed.
Entering text isn't the only thing I do, however. I already have a nice folding keyboard for when I want to enter text on a small device, so the most important thing is the display. Currently, I have two handheld devices with very different screen technologies. One has a thin-film transistor (TFT) screen. At 200 dpi, it's superb for video and dynamic user interfaces, and okay for text. The other screen is a 166 dpi E Ink display, which is gorgeous for text and monochrome illustrations, and terrible for anything else. I can read the E Ink display for hours without experiencing any problems with my eyes, but the TFT is not so good, and doesn't work at all in direct sunlight.
TFTs are starting to see competition from organic light-emitting diodes (OLEDs), which are very different from liquid crystal displays (LCDs). An LCD display relies on a bright light behind it, and blocks the bits of light the user shouldn't see. An OLED emits light the user should see, which means that OLEDs can be made transparent. I expect the next generation of mobile displays to be two layers: OLED on top, color E Ink underneath. If you can get the response time for E Ink up to about half a second, that's enough for entering text; I've used SSH over connections with more latency than that, and you only notice it at the start. The current response times, around 0.9 seconds, are just a bit too long. This presents some interesting challenges for GUI designers. In most applications, I'd expect to see the document drawn on the E Ink display, and floating palettes and other dynamic UI elements drawn on the OLED. You might see some things drawn in a two-stage way, so that characters you type or lines you draw are first shown on the OLED and then on the E Ink display.