For the first version, we want to implement lots of extractors.
Handlers will be fairly simple, as we don't want to spend time on extracting metadata during chunk calculation yet.