4:03:10nyefSo, careful system design should be able to handle swapping out device hardware across a checkpoint/resume. That mostly leaves the need to clear out corrupted state of some sort as a reason for a clean boot?
4:06:03nyefBut corrupted state IS a reason to reboot... and persistent corrupted state is a known weakness of orthogonal persistence systems.
4:07:02nyefIs there a "non-corrupted-state" reason to reboot?
4:07:24beachIf the file system is corrupted, then there is the same problem with traditional systems.
4:07:57beachI see no reason why one could not have a "fsck" for a system with orthogonal persistence.
4:09:45nyefThe intutition I have about that is that "fsck" basically has to deal with one (middling-complex) data structure, while doing so to an orthogonal persistence system has to deal with every piece of state in the system.
4:11:09nyef... Right, fsck doesn't help with corrupted files within the filesystem, you end up needing other mechanisms for that.
4:11:48nyefWhich is why you have backups and the like.