With Garbage Collection coming to CernVM-FS, we need to take extra care of getting all internal object references right. Thus, the history database of named snapshots needs some additions and audits.
These are things from the top of my head:
- maintain a previous_revision link to older history database objects
- allow for hidden named snapshots
- instead of deleting a tag hide it
- delete only if the garbage collection ran and removed the snapshot
- keep (hidden) references to all snapshots, even if CVMFS_AUTO_TAG is off
- needed to keep historic revisions reachable even without a continuous root catalog chain
- re-think the management of named snapshots in the CLI
- i.e. through a dedicated cvmfs_server command
- maybe allow management of named snapshots without a re-publish of the root catalog?
- recreate a (dense) history database from a legacy repository
- potentially a cvmfs_server migrate step in 2.1.20