This page describes some general guidelines for both reviewers and authors. When justified, there may always be exceptions to these guidelines. But authors be forewarned, you may not get an lgtm on patches that don't adhere to these.
All benchmarks and unittests must work for all Chromes from the current stable version to tip of tree. Where new Telemetry features rely on new Chrome features, the benchmark should gracefully degrade. Backwards compatibility fallbacks may be cleaned up after they are no longer necessary for the stable channel.
All features must be generalized to have a reasonable path to support on every platform where Chrome runs (Android, CrOS, Linux, Mac, Windows and soon iOS).
We aim to take advantage of the best support on each platform, not conform to the lowest common denominator across platforms. So some APIs may indicate that a give functionality is not supported for the platform.
Since it isn't always feasible to implement all platforms in a single patch, some platforms may raise a NotImplementedError. In this case, a crbug should be filed for the platform implementation and reasonable effort should be made to implement or find an owner.
Telemetry must work for our buildbots and chromium developers without requiring a compilation or manual installation/configuration. This means assuming no more than a Chrome checkout and python 2.7.
In cases where native code is necessary, we check in prebuilt binaries for all relevant platforms to cloud storage. In cases where installation/configuration is required, Telemetry scripts it for the user.
Code must adhere to the Google Python Style Guide.
All features, large or small, must have unittest coverage. We welcome new unittests, especially where there's no current coverage! Your patch/feature will be much easier to accept should you decide to first bring some test coverage to the existing code, and then add your new feature.