Loading ...
Sorry, an error occurred while loading the content.

6190Re: [decentralization] decentralized recognizability

Expand Messages
  • Lucas Gonze
    Nov 5, 2002
      On Tue, 5 Nov 2002, Miles Sabin wrote:

      > Lucas Gonze wrote,
      > > It seems to me that the technique itself could be refined and
      > > extended. For example, break the signatures up into several chunks,
      > > use color, modify the generator to output well known shapes. Also,
      > > there should be unix console utility to do it, apart from the circle
      > > implementation.
      >
      > At first glance looks quite nifty.
      >
      > But I'm not sure I know what the application is.

      The thing that strikes me about it is that it's a way of factoring
      recognizability out of the problem of human-manageable long numbers. The
      application is to eventually be part of a complete solution.

      > What do you gain by
      > having an only moderately reliable way of seeing that an agent holds a
      > particular public key? And surely it's not that hard to generate a
      > distinct public key which is sufficiently visually similar to the
      > expected key to be easily confusable with it?

      I imagine it's a lot easier to generate a usable spoof of the image than
      to break the key itself. But the key itself is also moderately spoofable,
      in the sense that keys can also be visually similar ("l" instead of "1").
      The question is whether it's easier to do a visual spoof with bezier
      curves than letters.

      > My gut reaction is that refining and extending the idea in the ways
      > you've suggested might not be all that productive ... possibly counter-
      > productive even. At a wild, uninformed, guess I'd say that colours
      > might drown out information in the squiggle rather than augment it.
      > Perhaps if you're using a handwritten signature metaphor, then keeping
      > things monochromatically familiar would help with recognition.

      I'm thinking of colors as an aid to chunking, because people appear to be
      able to remember more stuff if it's split up among multiple axes.
      (Assuming that the ancient paper I read on this, the George A. Miller "The
      Magic Number 7 plus or minus 2" 1956 classic, is still credible).

      > On the face of it, this'd be a good research project for someone who is,
      > or knows, a cognitive psychologist or human factors person.

      Yup.

      - Lucas
    • Show all 17 messages in this topic