dump works at the device level, dumping the raw block device by interpreting the ext2/3 structures there. If you pass it a directory, it converts it to the device mounted there and dumps the device. restore, on the other hand, operates at the filesystem level. You don't need to be root to dump. Your dump script can run as anyone in the "disk" group, the default group of disk block devices, which by default have group read access. You do need to be root to verify, though, because restore is going through the filesystem. I back my CentOS box up to a USB-attached hard drive on a Windows XP workstation mounted via cifs. This is effectively a "push" system. After the backup, I run "restore -C" to verify that the data got there successfully. During the verify pass, I remount the filesystem with the noatime option so that reading it to compare to the "tape" image on the USB drive does not change the atimes. I then re-enable atime when the verify is done. (I use atime to watch for dead email accounts and so that tmpwatch will work correctly.) dump has its own home page and mailing list, and the author is very helpful with support. http://dump.sf.net/