bbengfort
commented
7 years ago Right now each object name has it's own version, represented as a lamport scalar version number.
The namespace maps: global name --> version number. The version bucket maps version number --> file metadata
The version number is currently composed as a tuple:
(global name, scalar, process id)
The global name is required so that (test.txt, 1, 1) isn't confused with (foo.txt, 1, 1) since each object owns it's own history.
On Flush, a new version needs to be created, and the previous version number needs to be stored. This new version needs to be flushed to disk.
Create the ability to lookup historical version information from FluidFS.