It’s My (Third) Party, and I’ll Cry if I Want To
summary
Shows how to measure actual cost of third parties and then fix them: use WebPageTest + RequestMap to map domains and run block-tests, use Charles throttling to simulate slow origins, and inspect runtime where a tag manager can be 25% CPU or more. Presents concrete impacts — 200 ms of third‑party latency equates to millions in lost revenue and Walmart reports 100 ms → ~1% revenue — and concrete fixes: self‑host, async/defer, preconnect/resource hints, CSP, SRI, priority hints.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 7:56Security
- 11:12Delays
- 14:05Outage
- 16:47Risks
- 17:07Request Map
- 21:56Charles
- 23:59Missing Files
- 27:07Runtime Cost
- 28:47Discuss
- 30:21Advice
- 32:19Internal Teams
- 35:00SelfHost
- 37:12Content Security Policy
- 38:28Resource Integrity
- 39:18Priority Hints
related talks
47mIt’s My (Third) Party, and I’ll Cry if I Want To
50mIt’s My (Third) Party, and I’ll Cry if I Want To
31mIt’s My (Third) Party and I'll Cry If I Want To
47mIt's My (Third) Party and I'll Cry If I Want To
52mIt’s My (Third) Party, and I’ll Cry if I Want To
46mChaperones and curfews: minimising 3rd party impact
46mDeep dive into third-party performance
13m"Don't Let Third Parties Slow You Down"