The output file uses variable-length integer encoding (often referred to as "varint" or "ofs_int" in XDelta documentation). This allows the file to represent small numbers (like short copy distances) in just 1 or 2 bytes, while larger offsets use more bytes, optimizing the size of the metadata.
The body of the output file is divided into "Windows." A window represents a contiguous section of the target file. output file xdelta