- remove practically never used encodings (UTF-32, ARMSCII, Latin-4...),
 - remove duplicates and/or subsets (UCS, Latin-1, TIS620...) that I know of,
 - provide human-readable names (fixes #2571).