Polling: It's Good Enough for the WWW & It's Good Enough for You
summary
Argues that polling often wins on cost and scalability: push scales linearly with subscribers while polling plus caching can make many clients hit a single cached response. Gives a numeric Twitter thought experiment (300 push requests versus at most 200 cheap polls), points to HTTP conditional GET and 304 Not Modified as concrete tools, discusses pubhub subub flow and cache/memcache strategies, and reserves push for very low‑latency, high‑frequency cases.
AI-written summary · about
chapters (13)
- 0:00Polling introduction
- 1:56Push vs poll styles
- 3:01Polling history
- 4:07Why push is expensive
- 5:54Polling scales better
- 7:01HTTP conditional GET
- 8:05Push is complex
- 8:48PubSubHubbub
- 10:13Push vs polling
- 11:17Push vs poll choice
- 13:27Live music & Q&A
- 14:11Efficient pushing with tags
- 15:13Q&A: Mobile notifications
related talks
28mThe End of Polling
31mFrom Polling to Pushing: Going Realtime with Go, gRPC, and Redis
58mSSE vs WebSockets vs Long Polling
26mThe end of polling
57mWhy and how to transform a REST API into a Data Streaming API
53mComet, Ajax, web browsers limitations and web apps performance factors
38mProcrastination as a Service: Caching and Queueing
21mHigh Performance APIs through Caching