> Well, the problem is that while I do occasionally need cyrillic chars,
> I also need English, Estonian, Finnish/Swedish, Latvian and Lithuanian.
Probably this is the first example ever appeared on this list for the
demand of i18n database, that should be the future direction of
PostgreSQL, in my opinion.
Currently MB has two encodings for this kind of purpose: Unicode and
mule-internal-code. Both of them allows mixed languages even in the
same column. This might give you a partial solution.
> The only two of them that don't have overlapping character codes are
> Russian (all chars >127) and English (all < 128)
>
> My current solution is to run without --enable-locale and do all the
> sorting
> in the client. But it would be often useful to have language specific
> columns.
This is another reason why we cannot rely on the locale mechanism
supported by os. Basically locale is a global data for the entire
process. If each column has different language, we have to switch
locales. That would be inefficient and painful.
---
Tatsuo Ishii