Temporal databases permit retrieval of past instances of data, but currently available indexing techniques mainly focus on partial or total persistence and fail to efficiently deal with retroactive changes. Moreover, bulk loading and updating remain I/O-costly in large systems. In this paper, we present the Retroactive Buffer Tree (RB \(_f\) T), an external memory data structure that enables retroactive insertion, deletion, and update through buffer tree-based batch updates coupled with version-sensitive indexing. Unlike other multi-version B-tree variants, the proposed technique permits retroactive updates without preventing future changes from being applied to multiple versions efficiently using partial persistence. A theoretical analysis along with an experimental study of the proposed technique focuses on I/O complexity, efficient bulk loading, and space requirements. Experiments on a synthetic workload reveal significant reductions in I/O costs compared to existing multi-version indexing techniques under heavy workloads. This work is restricted to the algorithmic study of retroactive indexing structures and does not incorporate full integration into database management systems.
