Btrfs was written to do things that ext4 and XFS structurally cannot. Scale further, verify its own integrity, and manage volumes itself instead of leaving that to a layer underneath. It is copy-on-write throughout, keeps both metadata and data in btrees, and checksums both.

I went through the design, then the implementation, then the feature list as it stood at the beginning of 2012: snapshots, checksumming, scrubbing and built-in volume management. On paper it is the file system everyone said they wanted.

Then the part that mattered to anyone thinking about production. Development was very active, but the file system checker was incomplete and RAID 5/6 was not finished either. Those are the two things you need most on the day something has already gone wrong.