I want backup software that will keep track of my files and when they were last used (NTFS last accessed timestamps). After a certain threshold, the files would be backed up to another location, BUT, and here's the imp bit - a stub would be left on the filesystem that would look like the real file to the OS, and thus to any program. When someone tries to access the file, it could either be read from the backup location (if online) or the user would be prompted to connect to the correct media.
I should be able to define folders and time limits for specific filetypes, so I can say that anything I haven't seen in 4 months in c:\Movies can be archived, but My Documents should always be available.
Another imp feature I find missing from backup software is that they don't track media. External USB drives and optical media both have unique id's, so instead of asking me to insert drive K: the program should be able to detect if I attach the right drive, no matter what drive letter is assigned.
So I'd be able to say that I want all my movies and music backed up to my 250GB usb drive. The program would keep track of what needs to be done, and whenever I happen to connect it, the files would be backed up.
I know backup like this exists for corporate use across multiple platforms in SAN's, but haven't seen anything for the home user. IMO, combined with CDP (continuous data protection), like what FileHamster provides, this is what's needed for backup to become mainstream and not just be used by tech savvy people. Instead, backup programs today present a bewildering array of choices - full or incremental, where to backup, how often to do it, proprietary formats, compression levels etc. No wonder most people don't use them.
Backup software should detect if I have extra space available (on a different partition, disk, NAS, whatever), ask me if I want to use it as backup location, and never bother me again. We now have revision tracking built in to the OS (Vista's Previous versions, Apple's Time machine), there's no reason not to have automated backups that are set-and-forget.
Am I being overly optimistic or does something like this exist?
Edit - the motivation behind this is the philosophy of 'throw away anything you haven't used for a year', or at least put it away in storage! And this is the concept on which modern computing is built - virtual memory, paging, on demand loading, smart pointers are all implementations of this idea. So as a programmer, I can't help but feel this way