Brian Burg

Timelapse: back-in-time debugging for the web

Current web browser debuggers are inadequate for today's complex, event-driven, and nondeterministic web applications. The Timelapse project adapts back-in-time debugging techniques to the domain of client-side web programming. Unlike traditional breakpoint debuggers, Timelapse creates an exact recording of a web program's execution. This recording can be then played forward and backward and freely inspected at any time, using existing debugging tools. Timelapse also introduces novel trace analyses that help developers navigate execution recordings more efficiently.

Timelapse is currently under development. I am looking to collaborate with strong UW undergraduates who are interested in debugging, web applications, and/or dynamic analysis. Contact me for details.

Timelapse currently extends the debugger tools and runtime of the WebKit platform. More information is available on the Timelapse project page.

Collaborative Optimization

Recent advances in just-in-time compilation for JavaScript have made it possible to deploy large-scale applications using the HTML 5 platform. Unfortunately, web applications are still orders of magnitude slower than native applications. We propose collaborative optimization as a way to transparently improve performance by harnessing the "collective knowledge" about how individual web applications run. In essence, we extend traditional profile-guided optimization to collect profiles over many users, summarize profiles at web-scale, and then distribute optimization hints to any users that are able to take advantage of such hints.

Collabopt is currently under development; Once again, I'm looking to collaborate with strong students interested in browsers, language implementation/optimization, and data mining. Contact me for details.

I spent some time developing collabopt infrastructure while at Mozilla Research. For research updates, watch my blog!

Past Projects

Dynamics of JavaScript

JavaScript has recently become of interest to language researchers; these brave souls have targeted it for language extensions, program analysis, and other research. Depending on whom you ask, JavaScript is either well-behaved or extremely unpredictable. This is worrying, since most research must make simplifying assumptions: if these assumptions do not hold, then research claims may be unfounded. We aim to quantify how JavaScript code actually behaves in the wild, and whether this aligns with assumptions found in programmer folklore and the literature.

The DynJS project analyzes the dynamic (runtime) behavior of JavaScript in an effort to better quantify how the language is used. To date, we have focused on aspects of dynamicity in general (PLDI 2010), as well as uses of eval specificially (ECOOP 2011).

C3 - Lowering the barrier to browser extensibility

In areas of research that require software it is common for researchers to experiment and extend research software via runtime extensions or by hacking and recompiling project sources. Neither of these are palatable for web researchers: most web browsers consist of mountains of highly-optimized C++ code, and provide only a few ad-hoc extension mechanisms.

I helped build C3, an experimental HTML platform for web-related research, during an internship at Microsoft Research. C3 is built from the ground-up for flexibility: it is written in managed C#, it generalizes several existing extension mechanisms and adds new extension points in a systematic way. Several architectural features encourage modularity and experimentation.