This library has some rough spots, mainly around:

1) Usefulness: The library is not specially useful outside of
   tracker. In the source tree we already have plenty of extract
   modules that handle most popular file formats. Future-wise, if
   a brand new file format earns popularity, we'll want it in the
   tracker tree, and not as an external module. So neither
   implementing minor third party out-of-tree modules nor offering
   the possibility to reimplement a type Tracker knows well is a
   compelling thing to support.

2) Maintainability: The API is highly inconsistent, hardly
   introspectable and with little forward-compatibility thinking.

So in short, libtracker-extract is something we want to keep under
the rug, only available to tracker-miner-fs and tracker-extract.