I can see putting parts of NetBSD, Linux or other Unix in separate partitions,
so as to make upgrading simpler, without having to get in the way of user data
or software that is to remain unchanged.  One might want to try NetBSD current
while still retaining the latest stable version, and not have to install Emacs,
Lynx, gcc and everything else in duplicate.  But I've wondered why Linux prefers
a separate swap partition, considering primary partitions are limited.

OS/2 uses a swap file, SWAPPER.DAT, whose size is always an integer multiple of
1048576 (which is 2 ** 20) bytes, never a swap partition, and I think the
various versions of MS-Windows use something very similar.