In current system, there is no need for the commit to be unique. Commit are strings which could be anything. When a filesystem is versioned, the commit string is used on the relationships to denote the new change.
However, with rollback introduced, it would be desirable to have the commit unique so that given a commit id, we could obtain all the operations that belong to that commit and can negate those to produce a new commit with inverse operations (called rollback). If the commit id is not unique then there is no way to uniquely identify the operations and which commit they actually belong to.
In current system, there is no need for the commit to be unique. Commit are strings which could be anything. When a filesystem is versioned, the commit string is used on the relationships to denote the new change.
However, with rollback introduced, it would be desirable to have the commit unique so that given a commit id, we could obtain all the operations that belong to that commit and can negate those to produce a new commit with inverse operations (called rollback). If the commit id is not unique then there is no way to uniquely identify the operations and which commit they actually belong to.