A compact representation for file versions: a preliminary report

Abstract

A system is presented for the compact representation of multiple versions of a file. The presentation is in terms of vectors and matrices, which results in conceptual simplicity. Algebraic transformations enable the retrieval process to be optimized for any given version or set of versions, in contrast to always optimizing for the most recent or least recent version. Moreover, any version can be added or deleted without affecting any other. File differencing and dictionary compaction are unified, and data compression can be included. A compact representation for the (sparse) matrices is presented, and the main algorithms are described in terms of this representation.<>