# Live analytics script transfer experiment

Protocol recorded before measurements, 13 October 2026.

Question: How many response-body bytes does the public LiteStats tracking script transfer under different HTTP compression negotiations?

Target: https://getlitestats.com/track.js

Design: 15 rounds, four requests per round (60 GET requests total), sequential requests, deterministic seeded shuffled order within each round. Conditions: identity; gzip; br; gzip, deflate, br. Record actual Content-Encoding rather than assume the requested encoding is used. Each request uses a fresh HTTP/1.1 connection and no conditional cache headers. No page is loaded and no analytics collection endpoint is called.

Primary outcomes: response-body bytes, decoded JavaScript bytes, decoded SHA-256, status, observed content encoding. These are body sizes, excluding HTTP headers, framing, TLS overhead, and subsequent tracking traffic.

Secondary observations: time to response headers and response completion, measured from request creation; these include client/network/connection/server effects, and are descriptive observations from one machine, not page speed or browser performance measurements.

Validation: retain every attempt and failure; require HTTP 200 and a consistent decoded SHA-256; inspect observed encodings; compare the fetched source with the local tracking file. No additional rounds based on favourable results.

Record Node version, OS, architecture, timestamps and selected response headers. No claim of a geographic location inferred from client timezone.

Compute medians and ranges without statistical significance claims. Local gzip and Brotli compression examples, if reported, must be explicitly labelled as locally recompressed reference sizes, not live transfer sizes. Express kB as bytes / 1000 and KiB as bytes / 1024.

Publish as a vendor-authored, reproducible case study with raw data, source hash, limitations and reference links. No ranking, conversion, Core Web Vitals or competitor superiority claims follow from this experiment.