There are stable implementations for Linux, Mac OS X, Solaris, and BSD (at least). You can export a filesystem from any one of these, unplug the hard drive, plug it into another computer, import the filesystem, and go. Doesn't even have to have the same CPU endianness.
Besides being designed from the ground up to be operating system and CPU endianness independent, it does a ton of other stuff that no other filesystem does. It's a total rethink of the entire filesystem / volume manager / RAID pool concept. All of these things are merged into ZFS, with a user interface that's both radically simpler and far more powerful than managing all of those things seperately.
You can easily add capacity to existing datasets (up to 256 quadrillion zettabytes, it's 128 bit!), do mirroring and striping at the filesystem level, and add hot standby drives. You can combine drives of different sizes into storage pools. You can do rolling drive size upgrades in the future to expand the filesystem size when drives become cheaper, without taking the filesystem offline. It has built-in pervasive checksumming and self-healing data corruption detection, there's a strong focus on data integrity despite flaky hardware, writes are atomic, you can take snapshots and roll back to previous filesystem states, there's an advanced adaptive memory caching system that's far beyond simple least-recently-used caches..
Is that not enough? How about built-in encryption, data compression with your choice of algorithms, deduplication of file data... You can add a small SSD drive to your huge traditional hard drive based dataset to act as an L2 cache for fast, durable writes, fully online maintenance and upgrades, builtin SMB and NFS sharing, you can stream updates...