"Dimitry Sibiryakov" <SD@...> wrote:

>
> >And this is what I don't like. Doesn't it require the INTL stuff be
> >installed on the client side?
>
> No. Client library does not need to care about collations and
> sorting.

Here I agree. Collations are server specifics.

> For conversion from utf-8 to application charset and back
> system routines can be used. Such as WideCharToMultiByte on Windows
> or iconv library on POSIX.

Well, this makes some sense then. But doesn't it mean that users loose any
ability to work with charsets explicitly? I'm not sure that automagic
conversions based on the desktop locale/charset fit all needs. Or do you
suggest this as a default encoding/decoding path (to simplify the minimal
client installation) with an optional ability to use the client INTL
libraries too?


Dmitry