>> Bad news, :(
>>
>> I have to use Turkish collate...
>
>Maybe not always?
>In my area we've got similar issues with Collate=German.
>What I do is still using Collate=Machine as default.
>And locale collate only for things like an Index On Lastname etc.,
>IOW, only those (rare) expressions where a locale order is actually required.
>In my opinion, "machine" is preferable for an Index On phone, for example, and also especially for techno indices like an integer primary key, etc.

You are completely right. Collating sequences other than machine are only for language-specific text manipulation and sorting/ordering; anything else should be done in machine sequence. Non-machine sequences are slower, using two bytes for each character, and are mapping multiple characters or character pairs into same byte pairs (in Hungarian, for example, cs, CS and Cs will be equivalent, and will sort after ct, cu, cv, cw, cx and cy, but probably before cz), which would give inexact searches for anything that's not a string in that language - a date, binary or numeric tag.