FileBuddy
Scientific Standards

Benchmark Methodology & Standards

Transparent, reproducible measurement principles governing FileBuddy’s empirical performance baselines and client-side execution benchmarks.

1. In-Memory Client-Side Execution

All benchmarks execute entirely within the user’s local browser runtime. Test fixtures are generated deterministically in memory as native WebAssembly / JavaScript buffers. Zero file bytes are transmitted across the internet to FileBuddy servers or third-party telemetry endpoints.

Because processing relies on the client device’s CPU architecture and available browser memory, absolute timing will naturally vary between desktop workstations, laptops, and mobile devices.

2. Synthetic Fixture Generation

To guarantee 100% privacy and legal cleanliness, benchmarks never use private or copyrighted user documents. All test files are synthetic fixtures generated on-the-fly with known structural parameters:

  • Vector PDF Fixtures: Multi-page A4 documents with structured Helvetica font streams, standard metadata tags, and vector coordinate text.
  • Raster Image Fixtures: Synthetic uncompressed RGB image buffers with defined dimensions.
  • Corrupted Document Fixtures: Intentionally damaged binary structures (e.g. prepended gateway junk bytes or missing OpenXML manifests) used to evaluate File Doctor repair accuracy.

3. Timing & Throughput Measurement

Timing is measured using the browser’s high-resolution timer API (performance.now()). Measurement begins immediately before passing the fixture buffer to the functional processing primitive and concludes upon the resolution of the output Blob.

Timing includes format parsing, transformation, memory allocation, and compression. It excludes artificial network delays or user interaction pauses.

4. Integrity Assertions & Quality Metrics

FileBuddy refuses to publish arbitrary, subjective "quality percentages". Every benchmark score represents an objectively verifiable boolean assertion or mathematical delta:

  • Byte Reduction Delta: ((InputBytes - OutputBytes) / InputBytes) * 100%.
  • Page Integrity: Verifying output PDF page count exactly matches expected document structure.
  • XML Schema Validation: Confirming OpenXML packages unzip cleanly and validate against standard schemas.
  • Magic Byte Matching: Confirming restored files begin with valid container signatures (e.g. %PDF- or PK\x03\x04).