Drupal is a fast, open source and highly customizable CMS used by nearly 800000 websites in 320 countries. It has this important feature that every user can easily create and customize a website in a way that fully meets his needs. In order to do that, Drupal counts more than 10000 different modules that are highly complicated to put together because of the number of dependencies between them. This situation makes Drupal incredibly hard to configure reliably.

This session has two goals:
1) show how to elicit, model and display the module dependency graph of the whole drupal platform in an efficient way
2) demonstrate how we can improve the configuration management of a drupal distribution

First, we will introduce new techniques coming from 5 years of software engineering research and centered around the TVL variability language. Next, we will show how we managed to extract the module dependency graph of drupal. Then, we will perform several analyses on this dependency model (like conflict detection or module suggestion). Finally, we will show how our techniques pave the way for automated configuration.