Arno,

>
> How about indexes (keys) that need to hold the exact collation you defined
> (case-insensitive / equal-characters in weight)? It is importand to store the
> key in the exact character repetoire or did i miss something in the proposal?

We've already got the problem of collations and indexes for Unicode-FSS
and other character sets that can support multiple collations - Latin-1
and the various French and Spanish collations come to mind. I think
that the ultimate solution is going to involve one character
representation and different indexes for different collations. Which
is probably better than having multiple character sets and different
indexes for different collations based on different character sets.


Regards,


Ann