On 2009-01-24, ABCD <en.a...@gmail.com> wrote: > There actually is a good reason (oddly enough) for Windows > using a file on the filesystem for its swap space. Because it > is a simple file on disk, if Windows realizes that the swap > file is almost full, it can expand your swap without having to > do things like repartition. This makes the "swap is full - > out of memory"-type problems less likely to occur
While that's a valid point in theory, I've never had a "swap is full - out of memory" problem in all the years I've been running Unixes that swapped to dedicated partitions. In my experience the system usually slows to a standstill and requires drastic action long before swap fills up. > (unless it is "filesystem is full" as well :) ). That, on the other hand, I do run into quite regularly. So it seems to me that using a swap file rather than a paritition is increasing the liklehood of problems rather than decreasing it while at the same time adding both system overhead and instability. Surely it's easier to corrupt a swapfile that's in a normal, heavily-used filesystem than it is to corrupt a dedicated swap partition? The code that prevents one partition from "spilling over" into another is much, much simpler and more bullet-proof than the code that manages blocks/clusters within a filesystems. If I were to guess why Windows doesn't use a swap partition, it would be because floppy disks didn't have partitions. -- Grant