Tika has a MIME type detector that is simpler and perhaps just as effective as the current Aperture detector. Since we're already dependent on Tika for text extraction, we can remove the Aperture dependency altogether.
We should also change how multiple MIME type detectors work together. Currently, the first detector to return a MIME type will always win, and Aperture always returns some MIME type (e.g., text/plain for text files or application/octet-stream for binary files). So rather than have the most capable detectors configured first, we should have the most specialized:
- Custom detectors (future enhancement, see MODE-xxxx)
- Extension-based detector that only knows about special cases (not as it currently defined with a complete 'mime.types' file).
- Tika-based detector
Note that we need to make sure to keep any extension in our existing 'mime.types' file that is not found in Tika.