Fixing Google’s Feedburner problem, and caching that half works
Google's Chrome Releases blog was experiencing stale results on its Atom feed, as discovered in a previous post. The CCBot, used by Opera, was parsing the feed to notify about Chromium vulnerabilities, but faced the problem of outdated results. The issue was resolved by discovering that the RSS feed could still be accessed with a secret parameter, ?redirect=false, which served fresh results.
The Feedburner Atom feed, which had previously refused caching techniques, started supporting them when accessed using the redirect=false feed. However, the service still didn't send an ETag, ignore If-Modified-Since, and tell everybody not to cache anything. The feed was served with cache-control: public, must-revalidate, proxy-revalidate, max-age=1, allowing a copy to be kept for one second.
By adding a cache-busting parameter to the URL, the cache was bypassed, and requests were served directly from blogger-renderd, ignoring If-Modified-Since.
The redirect=false feed responded with a 304 Not Modified sometimes, which was inconsistent. Moreover, If-None-Match worked on the redirect=false feed but only when requesting a compressed response (gzip). The full feed was served with a 1,725,860-byte uncompressed size and 57,817 bytes gzipped, making a 304 response very beneficial. Despite Feedburner still serving outdated results, the redirect=false feed solved the caching issue and allowed CCBot to function properly again.
Rewritten by urgent.news; may contain errors — check the original.
Source: joshua.hu