Four short links: 9 February 2010

Government Dashboard, Science Code Errors, Scaling Online Games, Information Theory

  1. Track DC — informative drill-down report from Washington DC government about the different departments. (via Sunlight Labs blog)
  2. Errors in Scientific Software — a 1994 study of scientific software that found inconsistent interfaces (1 in 7 for Fortran, 1 in 37 for C) and poor use of arithmetic such that significant figures declined from 6sf in the data to 1sf in the result. (via “If you’re going to do good science, release the computer code too” in the Guardian)
  3. How Farmville Scales — 75M players/month (28M/day), 1/4 of disk activity is writes, 50% higher load spikes, 3G/s traffic go between Farmville and Facebook at peak, LAMP stack, nagios+munin+puppet. (via Hacker News)
  4. Mathematical Philology — when two manuscripts of the same text differ, which is correct? This PLoSONE paper looked at all such discrepancies in Lucretius’s De Rerum Natura and found that the traditional principle of choosing the more difficult reading (on the grounds that errors are from humans unconsciously simplifying) has a strong information theory justification for it. Interesting to see this less than a week after an MIT Technology Review article on quantum teleportation remarked, There is a growing sense that the properties of the universe are best described not by the laws that govern matter but by the laws that govern information.
tags: , , , , , ,