Hi!
This is not CLDR, this is general collation algorithm.
http://cldr.unicode.org/index/cldr-spec/collation-guidelines
CLDR is a repository/process for LDMLs (thats what I referred to people sending us that data, in case current is wrong/not existing). Currently it has mistakes and multiple versions even for same locales - doesn't seem to be too stable nor correct.
An example: http://unicode.org/cldr/data/common/collation/lt.xml?rev=1.26&content-ty... ;-)
Do note, that such unstable changes require database rebuilds at each iteration.. So, we'd have to have someone reviewing it all, comparing with different sources, and then pushing it once every few years into some data staging environment where we do data conversions all the time? :) riiight...
Domas