One area which needs investigation is the method by which tasks will communicate with the runner. In AutoQA we did most of the output processing in each test and output a log through stdout. We want to take a more structured, scalable approach with taskbt by standardizing output from tasks so that the runner can parse that output easily.

The two choices we're aware of are [[ http://testanything.org/wiki/index.php/Main_Page | TAP ]] and [[ https://launchpad.net/subunit | SubUnit]]. Do an investigation of these two protocols and their supporting libraries. Obviously, the most important language for initial support is Python but we want to support other languages eventually.