Uhm, The standard Linux tool for handling software RAID devices has been around since 2001. And the LVM framework that lets you do NetApp-esque tricks like resizing volumes when you add disks to storage pools has been around even longer. (Granted, some of the filesystem-level support for the really fancy stuff, like growing volumes without reformatting, taking live snapshots, etc, is "only" about 8-10 years old, depending on which filesystem we're talking about.)
All this stuff has been in RedHat and friends (IE, the "industrial strength" Linux distributions) as out-of-the-box features for over a decade so I'm not clear what the heck you're talking about. Practically every multi-disk NAS device sold for home use based on Linux, which is most of them, uses software raid and LVM, only very expensive models have any hardware acceleration. And I've been using roll-your-own RAID 5s in home-built servers for that "over a decade" span. (I wouldn't recommend it for a desktop, but it works fine on a *file server* that doesn't need its CPU for much else.) I was going to assume you just meant "for desktop operating systems" but then you mentioned Solaris and FreeBSD, so... ?
It's also probably also worth noting that most cheap "RAID cards" or the "RAID" function built into many motherboards are also actually software RAIDs. *Occasionally* they'll have some rudimentary acceleration that helps distribute reads/writes when run in the simplest RAID0/1 modes, but usually they're just a plain IDE/SATA controller with a little BIOS ROM that provides an INT13H driver sufficient for recognizing a "This is part of a raid set" header on drives associated with a RAID container, "assembling" them into a single virtual device, and making it look like single drive in Real Mode so you can use it as a boot device. The Windows (or whatever) driver for the card needs to take over the software RAID-iness once it's booted. In that form Windows has supported "Software Raid" for positively forever. (Back to the 9x versions. The old Highpoint HPT370 chipset was one of the original poster children for this sort of device.)
Anyway, re this:
That is exactly correct. Strictly speaking a "Backup" should consist of disks/tapes/whatever that are only accessed to write the copy of the data and read it back in case of a disaster. (Ideally of course you have at least two sets, so if disaster strikes in the middle of you *taking* a backup that destroys both the original and the backup you're overwriting you have the last set to work from.) If your "backup" is online and writable when your computer gets infected with the Gawdknowswhat Worm and it decides to overwrite all your data with a billion copies of a bad photoshop of Mr. Ed with Britney Spear's face it's just as gone as if you didn't have it "backed up" or on "redundant storage".