These entries give the correct answers on a number of topics.
I asked a computer shop for a quote about a PC configuration featuring 4GiB of ECC RAM. The shop gave me a quote, but omitting the ECC part. When I pointed out the omission, I got an answer stating that ECC is not needed unless you want to do "very precise calculations".
Update on 2010-02-08: Also read the follow-up .
In my previous post "On the need to use error-correcting memory" I used per-bit upset rates given in Eugene Normand's paper to present-day memory sizes to obtain a likely bit-error-per-three-days figure for 4 gigabytes of RAM. My aim was to make a rational decision about using ECC memory or not. It has been pointed out that this computation is invalid since it applies to old memory chips which had much lower densities. Strangely, it is quite difficult to find reliable figures about modern DRAM chips so I re-did some calculations. Also, partial data released from Google indicates that you have about one chance in 3 to have a computer that gives much more memory errors than what my initial computation gives - but this time, not because of atmospheric neutrons, but because of bad hardware. Still, ECC helps even in that case.
The basis for rationality is acceptance of an external objective reality. Science attempts to produce knowledge that is as universal and objective as possible within the realm of human understanding. Traditionally, such knowledge was expressed thru language, preferably written language. Today, scientific knowledge can be expressed in electronic form: data sets, models and algorithms are increasingly published as results of the scientific process to be shared with other researchers and the general public.
List of short opinions on various window managers, with a bias for tiling WMs.