David McCusker posted a very nice maze recently. I solved it this morning, and it took a gratifyingly long time - I'm used to solving mazes in my head.
Someone posted a link to Olin Shivers's maze page. Olin's and David's mazes are both mazes, but there the similarity ends. David's maze, hand-drawn, has a distinctly organic flavor. More than that, it has an architecture. The different locations in the maze are places. In Olin's mazes, they're just grid coordinates.
In any case, David posted a challenge to me, to come up with an image processing algorithm to represent images in the line widths. I think the right algorithm is Ostromoukhov's Artistic Screening. I don't have an implementation handy, but I faked something similar in the Gimp. The results weren't that satisfying - I was hoping for more crispness in the image. Here's the source image, for reference.
I fly tomorrow to meet the kids in Denver. Kinda surprising how much I miss them after only a couple of days apart.
I'm happy these days. I'll write up and post a bit about my philosophy of "humble elitism" which I think is partly responsible. An example will do for now: plain old elitism is reading Slashdot, being irritated at how stupid the comments are, and feeling all superior. Humble elitism is reading better blogs, such as Hack the Planet.
I spent a good part of the day going over potential references in my queue. Here are my conclusions:
The author presents a bad trust metric, shows that it is not attack resistant, and seems to imply that centralized indentity service (a la VeriSign) is needed. Obviously, I don't agree. He gets a brief mention in Chapter 1.
Interesting. Their community-finding algorithm is very similar to Advogato's, so much so that it really feels like convergent evolution. However, they're not trying to make it attack resistant, and, indeed, it's not. I wrote a page or so at the end of Chapter 3 describing the differences and their effects on attack resistance.
The paper they have up there is white, but with writing on the pages (see alancoxonachip). They present a trust metric in fairly vague terms, but there's no reason to believe it's any good. No cite.
The new text is, as always, available in the thesis snapshot.