Classifying Research Software

With apologies to Borges, I believe that all research software can be classified as follows:

1. That which works on the original author’s laptop.
2. That whose only documentation is a 2009 conference poster.
3. That which depends on a library that depends on a library that depends on Python 2.
4. That which is included in this classification.
5. That whose version number “final_FINAL_v3_revised_corrected” is accurate.
6. That which produces subtly different results on Debian and Ubuntu for reasons no one has investigated.
7. That which was written in Fortran and is therefore the most reliable of all.
8. That which works correctly only when the input file contains an even number of lines.
9. That which when run produce a single number which is cited in seventeen papers.
10. That which would take less time to rewrite than to understand.
11. That which has been rewritten four times and is still not finished.
12. That whose unit tests that all pass because the tests were written to match the output.
13. That whose output is so plausible its authors have not questioned its correctness.
14. That which, from a distance, appear to be maintained.

See also this post.