Ajax techniques have become popular because they enable developers to build more complex web apps and provide a richer user experience. However, in the rush to push the envelope of what’s possible to build, the question of what’s fast enough has largely been sidelined. In fact, most of the Ajax advice you’ll find today often comes with a surprisingly high cost when it comes to performance, both of the initial page-load and during subsequent user-interaction. Since users tend to care more about speed than flashy features, this trend makes it hard to ship large Ajax apps that are truly fast enough for mass consumption.

This talk will discuss much of the Ajax community’s common wisdom—from wiring event handlers to using frameworks like dojo to using OOP/MVC programming styles. For each technique it will explain the performance costs and suggest alternative strategies that are faster. Drawn from Plaxo’s experience building an ambitious new integrated online address book, calendar, and data-sync web app (released Spring 2007), this talk will also explain how to measure and optimize web app performance in general, with a focus on writing high-performance JavaScript code (performance issues like caching have been covered elsewhere). The talk will conclude by arguing that designing web apps is more like designing for mobile devices than for the desktop—the platform is flimsy and being stretched beyond its original intent. As a result, every design decision must consider the performance impact up front in order to avoid costly re-writes for performance optimization.