HTTP Compression Checker (gzip, brotli)

Two measurements of the same page: the uncompressed body size (fetched with automatic decoding, like a browser cache would store it) and the compressed body size (fetched with an explicit Accept-Encoding header and no decoding, counting bytes on the wire). Savings = 1 − compressed / uncompressed. That is the real number your users' traffic drops by — not a theoretical benchmark.

Both are fine; brotli compresses text roughly 15–20% tighter at comparable speed, and every modern browser supports it — but only over HTTPS (CDNs simply refuse brotli on plain HTTP). The practical recipe: enable brotli for text content and keep gzip as the fallback; images, video and archives are already compressed — do not waste CPU re-compressing them. The checker reports whichever encoding the server actually picked for a browser-like request.

Pages under 1 KB skip the second measuring pass: compressing a few hundred bytes saves less than the compression header itself weighs. For small pages the tool reports the encoding the server declares, without nagging about percentages — attention is directed at 32 KB+ responses where compression actually matters.

nginx: gzip on; gzip_types text/css application/javascript application/json image/svg+xml; (brotli via the ngx_brotli module: brotli on; brotli_types …). Apache: LoadModule deflate_module + AddOutputFilterByType DEFLATE …. Behind a CDN, enable compression in the CDN settings first — the origin's headers may be overridden at the edge. Re-run this check afterwards to verify the savings actually materialized.