I spend a big chunk of my time over at LiveJournal with friends, so I see most of all the quizzes that go by. The other day I saw Charles Miller's insidious plot, and was almost too quick to comment about the problems; being a software engineer I was anxious to find their solutions, and did. Tracking quiz answers the users assume are anonymous then deanonymizing them is scarily viable.
So today I see this on my friends list. Check out this image URL phatjoe would have me post:
http://www.phatjoe.com/mandelbit.cgi?xcnt=0.29999999999999999995663191310057982263970188796520233154296875000&ycnt=-0.20000000000000000004336808689942017736029811203479766845703125000&scale=4.31975450785830616950988769531250000000000000000000000000000000000&inside=052305&close=552cf5&far=e02bc8
This solves the unique ID problem, especially as there's a real technical reason for so much data in the URL. That the fractal formula arguments are included makes it look especially like he isn't recording or tracking data: if someone knowledgeable saw you put in 217007 and got a really randomish section of Mandelbrot--and especially the same section--he would realize there's a record of the generation. Any high data process like fractal generation that would let you deluge the URL with data would let you steganograph the ID (or just use the sparse points in argumentspace for the IDs).
I don't mean to say phatjoe is identifying LiveJournalers with this, only that it would be possible to use random fractal numbers as a burr (bugburr?) visitors might pick up and carry with them. The question then is how to require (something like) a mandelbit in every quiz result. Maybe as a decorative border, or background image? That might be a little suspicious, but it could work if that's the theme of the quiz site (say, fractalquiz.com). After all, we're only trying to get the people who won't blindly copy HTML code as it is.
Er, or would be, if we were really doing this. Yeah. All theoretical like.
