Steve Ackman writes:
In 97mVj.1864$T1.1192@trnddc01, on Sat, 10 May 2008 18:49:41 GMT, Jim
Chandler, wrote:
Edwards attributes that to a lucky twist: The computer was running an
ancient operating system, DOS, which does not scatter data all over
drives as other approaches do.
Um, hello? DOS has and needs defragmentation tools.
"Other approaches" don't.
If all you're doing is writing data to the disk, it does pretty much
start at one end and fill from there -- which is what I'd expect to
happen if you're recording experimental data. It's when you're doing
mixed creation and deletion that things start getting ugly.
Hmm... I'm trying to think of whether there's anything inherent in
the filesystem (i.e. the actual data structures on disk) that forces
this, or if it's just that the implementations use poor allocation
algorithms?