Leveraging Fine-grained Data Flows in Web Applications
James Mickens ( Harvard University )
- 14:00 9th June 2016 ( Trinity Term 2016 )Lecture Theatre B
A modern web page contains megabytes of HTML, CSS, images, and JavaScript. Loading such a page requires a browser to evaluate a complex dependency graph involving those resources; once the page is loaded, subsequent interactions between those resources and the user can lead to tricky-to-diagnose bugs. In this talk, I'll describe how tracking fine-grained data flows can allow us to reduce page load times by prioritizing the loads of the highest ancestors in the data flow graph. I'll also describe initial work in using data flows to assist with time-travel debugging (in which developers use a logging-and-replay framework to analyze buggy program executions).