It's the "indexing" and "background fragmentation" that do it.
M'Snot-NT appears to periodically "optimise" hard drive layout by shuffling stuff around.
What an awful thing to say to a person. :-) Seriously, that seems unlikely. I started out on mainframes and have really very little experience with operating systems that are not...
I built a testbed machine with Lose-2000, installed just the M'Snot, installed their latest patches, defragmented, disconnected it from the net and let it sit "idle" for a week. At the end of the week, I started the defragment analysis tool and it recommended a defragmentation.
With Lose-9x, I got the distinct impression that the system paid little heed to hardware errors that might occur. That becomes important when physical memory is low so it's always using its swapfile; on a slow-old hard drive. Occasional glitches in disc access that result on a core dump on Unixen, appear to be ignored. The corrupt data are "loaded", trusted and "used". Similarly when a write error occurs in "memory management", it seems to be ignored. What gets loaded back in is what the system believes to be what it thought it wrote. -- "Bernd Felsche - Innovative Reckoning, Perth, Western Australia ASCII ribbon campaign Economist E*con"o*mist-, n. X against HTML mail One with a ready explanation as to why and postings his last prediction was so wrong