Also, the backup system uses single instance storage, meaning that identical clusters on many pc's are stored only once on the server. In other words, a backup of 5 Windows XP directories will take up the space of 1 Windows XP directory.
Also, the disk mirroring does not require the same type of disk. They can be a mix of IDE, SATA, and USB, and of various sizes.
Oh, and it has Windows Media Connect.
That sounds really cool. But if you have a system which has a proper package manager, you just backup your manifest file or package database and restore from scratch. Debian .manifest files are typically several kilobytes per computer. I guess you couldn't do that with Windows XP/Vista clients though. Cool stuff though, I doubt you can do anything similar to the intellegent backuping with FreeNAS. Does WHS support shadow copies for the file sharing?
Interesting. But does the .manifest restore also restore user's settings and files, or do you need to back those up separately?
WHS supports shadow copies, yeah.
WHS goes beyond that... WHS does complete PC backups of every single hard drive, every single file, every single setting. Install something you shouldn't on your PC today? Restore to the state from yesterday, last thursday, or 3 months ago.
Think of it like making a Ghost image of your entire system daily... and being a heck of a lot more efficient at disk usage because with WHS after your first full backup, you are just saving the diffs.
I am not so fond of using images or image diffs for backup purposes. I understand they are mostly needed under Windows, but I rather just store the "metadata" of my systems, which on some operating systems is enough to bring it up to it's former condition. At least when stuff is owned by a specific package or module.
They're not actual images. They're collections of clusters.
Anyway, I'm not sure how your method works, but from the user's perspective, I'm not sure if it matters. The user just reboots with the recovery disk, selects the day to which he wants his computer restored, and is done. Wether it works with files, clusters, metadata, or magic goblins isn't a concern for the user.
As far as I can tell, both methods are, from a user's point of view, equally effective.