Linux EXPRES interviewed me at DevConf.cz 2014 about working on ext4 in Red Hat’s Brno office. The question they kept coming back to was why ext4 keeps getting new features instead of turning into ext5 at some point.

Because the features do not require it. Bigalloc and metadata checksumming are both worth having, but neither one changes the on-disk format enough to justify starting a new file system, and starting a new file system costs far more than people assume.

We talked about the alternatives too. Btrfs is not trying to beat ext4 on performance, it is after copy-on-write and snapshots, which is a different goal and should be judged as one. F2FS has the harder problem: it is designed for flash, and flash manufacturers will not tell you much about what their devices actually do inside.

A file system takes ten years or more to stabilize. That is the entire argument for ext4 in production, and it is not an exciting one.