There is a problem with sorting of the space in the cs_CZ locale (as in many other locales). According to Czech Standard 97 6030 Alphabetical ordering (Czech Institute of Standards, Prague 1993) [Czech: ČSN 97 6030 Abecední řazení (Český normalizační institut, Praha 1993)]: The space between two contextual characters should be considered as a single character. The space is sorted before the first letter of the alphabet. For example: Novak Zdenek Novakova Jana So I propose to change the behavior of sorting the space from <U0020> IGNORE;IGNORE;IGNORE;<U0020> to <U0020> <U0020>;IGNORE;<U0020>;<U0020> The problem is that this change takes spaces on the beginning of a line into account, which is not correct as it sorts "Novák" and "(space)Zounar" as (space)Zounar Novák instead of correct Novák (space)Zounar The same applies to multiple spaces, they should be considered as one space, so "Novák(space)Jan" and "Novák(space)(space)Zdenek" are incorrectly sorted as Novák(space)(space)Zdenek Novák(space)Jan insted of correct Novák(space)Jan Novák(space)(space)Zdenek Is it possible to fix this behavior in locale definition? It should definitely solve the problem and fulfill the standard.
There is no way to express "not at the beginning of a line". Therefore whatever way it is done there is a problem. There is no way to make everyone happy. I suggest to leave it as is. If you can get the previous authors of changes to the locale (see the file) to agree with a change I'll reconsider.
I contacted authors of the original cs_CZ locale, they redirected me to report it as a bug, which I did. So, let it be as it is. I'll change the locales myself on the servers where I need correct space ordering. If there is a change in possibilities of definition so it'd be possible to define behaviour of white space chars on the beginning or end of the text and repeated white space chars, it would be great. I have no idea how much work it would be to change the code and if it would be useful for anyone else.
(In reply to comment #2) > I have no idea how much work it would be to change the code and if it would be > useful for anyone else. It simply isn't possible with the current interfaces.