>>718692259
I'm wondering if df has seen some improvements in recent years wrt ZFS. Still, I would not rely on it. Should you end up using snapshots (and you should be) or clone datasets the water will muddy rather quickly. As for compression, any data that can be compressed should, as CPUs are fast enough that the decompression time is less than the I/O time spent transferring the uncompressed file.

>I'll never fill that much storage up
You're 30TB deep and you haven't learned yet?