For the search collators, we have a base set of rules in root, then for a number of languages those rules are used in conjunction with a few language-specific additional rules. The common part should be imported if possible, instead of being duplicated in each of the language collations that uses it. These are:
az ca cs da de es fi fo hr is kl ko nb nn se sk sv tr
Note that he also has a language-specific search collator that uses the base set of rules, but it modifies some of the standard rules as well as making additions.