Thursday, October 20, 2005

HTML Notes

I switched the main website from HTML 4.01 Transitional to Strict. I can't remember why I chose to use transitional in the first place. The only deprecated HTML I had was an align attribute in an image tag.

Students in the CS105 class I'm TAing have asked about indenting using straight HTML. This reminded me of a page describing hacks to indent using HTML. (I'd use CSS to do it, but we don't expect them to know CSS well.)

Saturday, October 15, 2005

Rubik's Revenge

I bought a Rubik's Revenge (a 4x4x4 Rubik's Cube) a long time ago. I did figure out a solution, but it's extremely slow. The 3x3x3 method I learnt when I was a kid is also fairly inefficient.

I was curious how people speed things up for both types of cubes. After browsing for a bit, I've decided the approach best for me is:
  1. Reduce to 3x3x3 case as described by Chris Hardwick. Now standard Rubik's Cube algorithms can be used.

  2. Solve the cross.

  3. Solve the first two layers except for one corner piece. Originally I had planned to follow the "ZB" system, which has you solve the first two layers except for one corner piece and one adjacent edge piece, but this leads to numerous cases for step 2 of the system. For now, I'd rather sacrifice a few moves to get one more edge piece in place to cut down the number of cases for the next step.

  4. Perform step 2 of the ZB system to put the last first-layer corner piece in place while at the same time orienting the last-layer edges. They are the last few cases on this page. (I don't think I could ever memorize the entire ZB system. I lose a few turns putting the last middle-layer edge in place, but I only have to learn a small fraction of the cases.)

  5. Either orient the last layer ("OLL"), and then permute the last layer ("PLL"), or for more efficiency, achieve both goals simultaneously.
One of the above sites also points to a site with videos of people posessing an amazing amount of digital dexterity.