Be lazy, but with strategy
summary
Argues that custom preloading beats PreloadAllModules for large apps and low-end or slow clients, and shows concrete techniques: detect network with navigator.connection, or use timeouts (examples: 500 ms fast / 3 s slow cleared from polyfills), read cores via hardwareConcurrency and userAgent, collect transitions in analytics and run a hidden Markov model (2D transition matrix + BehaviorSubject) to predict next modules. Reports improved perceived load but warns about data-collection overhead and timing getting out of sync when polyfills change.
AI-written summary · about
chapters (13)
- 0:00Start
- 2:37PERFORMANCE USING LAZY LOADING
- 3:46LAZY LOADING IN A NUTSHELL
- 5:36CUSTOM PRELOADING STRATEGY IN A NUTSHELL
- 9:14SLOW CONNECTION
- 16:53LOW END DEVICE
- 20:09USER BEHAVIOUR
- 21:06LINEAR USER NAVIGATION
- 22:47MACHINE LEARNING IN A NUTSHELL
- 26:36MARKOV CHAIN IN OUR CASE
- 27:51WHERE TO GET THE INFORMATION
- 28:34HOW TO USE THE MARKOV CHAIN
- 30:11CONCLUSION
related talks
17mBe lazy, but with strategy
21mPreload Strategies: Step in Time, Step in Time!
30mStep by Step Improving Startup Performance with Lazy Loading in Angular
41mNAS, Predictions, Preloading, Presudo Isomorphism
38mSecret Shortcuts of Loading Web Performance
21mThinking PRPL
32mImproving Startup Performance with Lazy Loading in Angular
26mWeb Performance in modern JavaScript world