http://stardict.sourceforge.net/StarDictFileFormat

{6}. The collation file's format.
StarDict-2.4.8 start to support collation, that sort the word
list by collate function. It will create collation file which
names .idx.clt and .syn.clt, the format is a little like offset
cache file:
First a utf-8 string terminated by '\0', then many 32-bits numbers as
the index that sorted by the collate function, they are not stored
in network byte order.
The string must begin with:
=====
StarDict's clt file
version=2.4.8
=====
Then two lines like this:
url=/usr/share/stardict/dic/stardict-somedict-2.4.2/somedict.idx
func=0
The second line should have a ending '\n' too.

StarDict support these collate functions currently:
{{{
typedef enum {
	UTF8_GENERAL_CI = 0,
	UTF8_UNICODE_CI,
	UTF8_BIN,
	UTF8_CZECH_CI,
	UTF8_DANISH_CI,
	UTF8_ESPERANTO_CI,
	UTF8_ESTONIAN_CI,
	UTF8_HUNGARIAN_CI,
	UTF8_ICELANDIC_CI,
	UTF8_LATVIAN_CI,
	UTF8_LITHUANIAN_CI,
	UTF8_PERSIAN_CI,
	UTF8_POLISH_CI,
	UTF8_ROMAN_CI,
	UTF8_ROMANIAN_CI,
	UTF8_SLOVAK_CI,
	UTF8_SLOVENIAN_CI,
	UTF8_SPANISH_CI,
	UTF8_SPANISH2_CI,
	UTF8_SWEDISH_CI,
	UTF8_TURKISH_CI,
	COLLATE_FUNC_NUMS
} CollateFunctions;
}}}
These UTF8_*_CI functions comes from MySQL in fact.

The file's locate path just like the .oft file.

Notice, for "somedict.idx.gz" file, the corresponding collation
file is somedict.idx.clt, but not somedict.idx.gz.clt, the
"url=" is somedict.idx, not somedict.idx.gz. So after you gzip
the .idx file, StarDict needn't create the .clt file again.