A Short History of Filesystems
conceptYou've met the names in format menus and error messages—FAT32, NTFS, ext4—and maybe wondered why there are so many, and whether the old ones are simply bad. The list looks like a parade of products, each replacing the last. Actually, it's a chain of disasters and answers: three generations, each built to survive the failure that ruined the one before.
- Simple ledgers (1970s–80s: FAT, early Unix filesystems). Fast and tiny. But lose power mid-write and the ledger itself ends up half-updated—scrambled records, orphaned files, a repair tool run at every boot, and sometimes real loss.
- Journaling (1990s: NTFS, ext3, XFS). Before touching the ledger, write a short note—a journal entry—saying what you're about to do. After a crash, replay the notes. The ledger stays honest. This generation still runs most of the world.
- Integrity (2000s onward: ZFS first at full scale, then Btrfs and Apple's APFS). Journaling protects the ledger from crashes—but it still trusts whatever the disk returns. This generation assumes the hardware itself can lie, and verifies everything.
You can date any filesystem you meet by which disaster it answers:
- FAT32 / exFAT (generation 1) still rule USB sticks and camera cards—not out of nostalgia, but because when every device must read every stick, simplicity is the feature.
- ext4, XFS, NTFS (generation 2) are the workhorses under most Linux and Windows machines today: journaled, fast, thoroughly debugged, and still fundamentally trusting of the disk.
- ZFS, Btrfs, APFS (generation 3) each took the new assumptions somewhere different: ZFS to servers and many disks, APFS to the SSD in your Mac, Btrfs into the mainline Linux kernel.
The practical takeaway is calm: the older generations aren't wrong or obsolete—each survives exactly where its trade-offs fit. But when a machine holds years of data that must stay verifiably intact, generation 3 is the one built for that job.
The fine print that sorts real understanding from a timeline:
- Journaling protects the ledger, not your data. Most journaling filesystems journal only the bookkeeping (metadata), not file contents—a crash can still leave a file's contents half-written even though the records are tidy. "Crash-consistent" and "your data is intact" are different promises, and the gap between them is exactly where generation 3 begins.
- The fsck ritual—the repair tool that checked whole disks at boot—shaped a generation of habits. Journaling made it rare; the copy-on-write design (next stop on this trail) made it structurally unnecessary: a ledger that is never edited in place can't be caught half-edited.
- Why generation 3 took until the 2000s: verifying everything costs memory and computation that 1980s hardware couldn't spare, and the silent-corruption problem only became visible at scale—when you hold enough terabytes for long enough, "the disk sometimes returns wrong bytes without any error" stops being folklore and starts being a statistic.