Agreed, and that's how zfs does it behind the scenes. When it detects a read error, it uses the pool redundancy (whether mirrors or RAIDZn, whichever you're usi

Agreed, and that's how zfs does it behind the scenes. When it detects a read error, it uses the pool redundancy (whether mirrors or RAIDZn, whichever you're using) to transparently recover from the error. Even if you set up a pool with only one disk you can still set it to keep multiple copies of all the data. A super-ZFS that automatically did that using remote mirrors would be interesting, but would also stretch the definition of "transparent" a bit.