The following text is copyright 2002 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.


An incremental world


By Scott Bradner

Network World, 09/16/02


I've been increasingly depressed about the sorry state of what passes for innovation in the network world these days. Technical "innovation" seems to almost always be some minor incremental "improvement" over existing technology instead of a new idea.


There may be some real innovation in areas like the creative accounting represented by Enron's off-the-books, debt-eating spinoffs. But I don't see many new ideas being presented to venture capital companies these days.


Every now and then I see something that seems truly new, and it reminds me that innovation is possible. What brought this to mind was a little article in Science News (which is, by the way, a very nice way to keep up with what's going on in the world of science) about Dasher. Dasher, described in the Aug. 22 issue of Nature , ( is a new way for disabled computer users to type with their eyes, but has much broader implications than that.


For quite a few years there has been technology that lets computer users who cannot use a keyboard type by looking at a matrix of characters on a screen. The technology tracks where the user is looking to figure out what character he is trying to type. This is an effective but laborious process and is prone to errors.


Dasher adds knowledge of the language being used to put up a display of the character selected and the most probable next characters for the user to select from. Once the user selects the wanted character pair, Dasher puts up the most probable next characters, until the word is complete. This is a much faster process than picking out individual characters because the list of most probable next characters is much smaller than the full alphabet.


This idea is particularly attractive to me because I used the same information about the frequency of one character following another in English in a computer program (which I called Homunculus) in the early 1980s in a museum show to print out random English-like words. But I would never have thought of this application of the information. I'm quite impressed.


Obviously, the same logic can be used to speed up the currently laborious data entry process on PDAs. It also should be used to speed up entry of languages with complex characters such as Chinese. Open source software to implement Dasher should be released early next year.


Innovation is possible, even if it is not all that common. Of course, there is another part to the problem; most venture capitalists don't seem to like new ideas. Innovation is scary because they do not have any existing return on investment framework to put it in, assuming that "return on investment" is a concept that the new venture capital firms understand after the pummeling of the last few years. I did have a venture capitalist ask me if I knew of any "quirky" early start-ups the other day, so maybe there is some hope on that front.


Disclaimer: Harvard has seen rather many things that seemed to be innovative - some actually made a difference. But the university has not expressed an opinion on this example.