Schneier on Security: The Future of Privacy
One of my friends used to rot13 the name of someone asking for information and make a plausable sounding name to provide. (He could do rot13 in his head.) This allowed him to get new junk mail and figure out who sold the information. It also served the purpose of poisoning these databases with a lot of invalid information.
A personal bookmark of sorts.
Spent a lot of time cycling during college. At one point a group from OCW (Orange County Wheelmen) was doing regular Wednesday evening training rides (during the summer, so it was still light). The group naturally tended to sort out by ability level, and I ended up mostly riding with a guy … whose name I had long forgotten. About the only things I remembered was that he was a very strong sprinter, raced, and claimed to hold the record for a solo ride across the country (somewhat of an odd combination).
Wandering around the web, ran across a reference in Ultracycling: UMCA Solo Transcontinental Cycling Records.
Transcontinental Solo Senior Man 17 Mar 73 - Paul Cornish, Tustin, CA SM - NY (presumably Santa Monica to New York) 13:05:20 / 09.38 (Time (dd:hh:mm) / Average Speed)
When we tried sprints – he would be almost instantly dozens of yards ahead. Shortly after he would burn out and be unable to maintain speed, and I’d catch him. Even burned out he could muster enough to tuck in behind me as I passed. Drafting he could maintain the same speed – so I couldn’t lose him. Somewhere between amusing and frustrating, that was … 
Have no idea what happened to the guy.
Insane – yet fun…
Someone posted a weblog entry (do not remember where) with a puzzle:
How to efficiently find a subsequence of values with the greatest sum within a larger sequence.
The simplest algorithm is to try all subsequences. This runs in time O(N^3), so is not useful except for very small sequences (say 100 or much less).
For some reason this problem stuck with me until I came up with an efficient O(N) algorithm. The basic notion is to first collapse runs like-signed to a single number. Next the sequence is scanned for triples that can be collapsed (either positive or negative), and the scan is repeated until no more triples can be combined. (Think of drops of condensation beading together.) Attached an archived Eclipse project containing the implementation and tests.
Watching a video about global warming, and this picture was flashed up as evidence.
EO Newsroom: New Images – Retreat of the Gangotri Glacier
Hold on … what does this prove?
Note that the area lost between 1780 and 1935 (155 years) looks to be about twice the area lost between 1935 and 2001 (66 years). Given that the amount of greenhouse gases generated by humans in the past century, what caused the larger retreat before 1935? If human generation of greenhouse gasses was the dominant influence, you might expect most of the loss would have been since 1935 rather than before.
If you were just looking at this glacier for evidence, you would have to conclude that “global warming” was already occurring (at least for this glacier) before human mass generation of greenhouse gasses. You would also have to conclude that humans have not had a big effect on the already-warming climate.
Of course, this is assuming there is a linear relation between temperature and the area lost by the glacier. At the very least, any change in slope of the valley probably could influence the rate of shrinkage.
So … what does this prove?
