A History of Fetching Specs
summary
Chronicle of how RubyGems moves from a full specs download to a tiny, CDN‑friendly index and whether that actually speeds installs and scales. It demonstrates a concrete new index format with names.list, versions.list and per‑gem depth files, explains operational pain points (uncacheable dependency API, heavy Postgres queries, single US origin), and reports real wins like cuts bundle install from ~18s to ~3s, depth folder ~200MB vs gemspecs ~1.1GB, and HTTP-cacheable static index served via CDN.
AI-written summary · about
chapters (7)
- 0:00Start
- 4:52Problem #1: No Dependency Info
- 7:04Dependency API
- 8:14Modern Index Problems
- 8:28MOAR RAM - Solved
- 9:19Problem #1: Not cacheable
- 12:25Problem #3: Operational Burdens








