Of course, disks in the Nineties were finite, as they are now
As I recall, this was not the case for Amoeba. Their hypothesis was that disk sizes grew faster than disk consumption and so a filesystem could focus on retention and skip deletion entirely.
Looking forward to the next one in this series, it’s when things start to get interesting. LFS really struggled with reads on spinning rust but with SSDs random seeking is fast and writing entire cells is fast, so most modern filesystems adopt some ideas from LFS. The next one will hit the period where journalling became ubiquitous (and fsck could finally die). It was also the start of the people starting to think about the places where layering should go, which became very apparent with ZFS. In FreeBSD, there were efforts to add journalling to FFS both in the filesystem and in the block layer. The former won in the end (though ZFS’s layering made the other choice), but the latter is still around because it lets you do fun things like have journaled FAT filesystems.
As I recall, this was not the case for Amoeba. Their hypothesis was that disk sizes grew faster than disk consumption and so a filesystem could focus on retention and skip deletion entirely.
Looking forward to the next one in this series, it’s when things start to get interesting. LFS really struggled with reads on spinning rust but with SSDs random seeking is fast and writing entire cells is fast, so most modern filesystems adopt some ideas from LFS. The next one will hit the period where journalling became ubiquitous (and fsck could finally die). It was also the start of the people starting to think about the places where layering should go, which became very apparent with ZFS. In FreeBSD, there were efforts to add journalling to FFS both in the filesystem and in the block layer. The former won in the end (though ZFS’s layering made the other choice), but the latter is still around because it lets you do fun things like have journaled FAT filesystems.