English: Subramanya Sastry and Gabriel Wicke will present details of how Parsoid handles the bidirectional conversion between wikitext and HTML. We'll present the requirements and challenges that make this both interesting and ugly at the same time. We'll present a high-level architecture of how Parsoid converts wikitext to HTML and then dive into some interesting details. We'll take some simple wikitext examples and walk through the stages of transforming the wikitext. Time-permitting, we'll also talk about our HTML to wikitext pipeline, and how we eliminate / minimize dirty diffs in that translation process.

In this talk, we will not be presenting the operational aspects of the Parsoid service as it is deployed on wikimedia sites (the cluster, caching, invalidation / update of cache entries, etc.). Beyond the 30-min presentation, questions and discussion will guide what we talk about. If anyone has any specific aspects of Parsoid they are curious about, please direct those to me and we'll try to include that.

We are going to assume basic knowledge of wikitext, templates, extensions, but will clarify as required. While this will be primarily a technical talk, we'll strive to make parts of this accessible in terms to those uninterested in all the gory details but nevertheless interested in the process.