 - make it faster!

 - interactive shell mode

 - compatibility:
   - install everything (config.ini, jmdict.sqlite) to package_datadir, to stop
     it from breaking when different paths are provided to setup.py install 

 - query types:
   - inflection!
     - and guess
     - if only there was Mecab for Python 3...
       - wrapper over binary?

   - whole-word for Japanese (feasible?)

   - support rōmaji regexpes
     - or at least warn about lack of support

   - perhaps it would be nice to make composite filters like search engines:
     - tea ceremony → matches any entry with both words "tea", "ceremoy"
     - "tea ceremony" → match exact phrase

 - data:
   - be sure we cover everything covered in EDICT 1, at the very least.
   - support lsource
   - clean abbreviations that are not being used
     - add our own symbols: ※、＊、(P)
   - more fine-grained --frequent (support ke_pri, re_pri)
   - support all JMdict languages

   - go beyond JMdict:
     - pitch-accent data (from where?)
     - names (JMnedict)?
     - famous people/place names from wikipedia etc?
     - order by frequency?

 - output:
   - make match coloring work for rōmaji output
   - better wrapping for "human" output
   - colored --help! (argparse extension?)
   - think hard about tsv output

 - better documentation
   - document tsv output
   - manpage?
   - about time to start a changelog...

 - config file support?
   - and option --saveconf

 - improve install script/distutils/updatedb
   - better handling of dev vs. installed environments
   - can we make pypi post-uninstall delete database?
   - deb?

 - localization
   - Japanese interface

 - unit tests

 - alternative backends?
  - better benchmarks to compare them
  - option to auto-compress/decompress database
  - backend: EDICT2 with grep
  - backend: some NoSQL thing?
  - backend: some kind of "full-text search" thing?
  - DICT protocol?
