Add some metadata in addition to the dataset's index and database file so that we can track (among other things) the current version of the dataset.
Additionally, once datasets have their version metadata, we can define a set of atomic upgrade strategies to upgrade from version A to version B. By adding an upgrade strategy from each previous version to each next version, it is transitively possible to have full backwards compatibility.
Add some metadata in addition to the dataset's index and database file so that we can track (among other things) the current version of the dataset.
Additionally, once datasets have their version metadata, we can define a set of atomic upgrade strategies to upgrade from version A to version B. By adding an upgrade strategy from each previous version to each next version, it is transitively possible to have full backwards compatibility.