I read some threads about set collate and now I'm scared (and sweating).
I use ONLY set collate to "czech" !
Maybe - I haven't problems with speed - because I have little tables (up to 500k rows) ? or 90% of indexes are on integer fields ? or ...
I try run some code under both sequences (with indexing) and the results are (for example): czech 1.8 second, machine 1.65 seconds .. (network traffic make more).
So, rewrite applications or not (many "order by") and can I mix machine indexes with czech collate for order by and (horror) can my applications now lies ?

Pavel