Roman,

> > But your reasononing is sound, especially in light of fact that
> > metadata BLOBs are marked as UNICODE_FSS, but store data in client
> > character set.
>
> Ok, then I keep using the client encoding. Would it be
> correct to write in the JayBird release notes that people
> should not rely on server converting BLOB contents to the
> client encoding and they better use either
> getBinaryStream() or getBytes() instead of getCharacterStream() or
> getString() for text blobs?

I think it would be correct to write that driver reads and writes blobs
using client character set always independently of their SUB_TYPE or
character set.
If user's database field character set matches client character set he
should be happy. If it doesn't user should be using getBytes and
getBinaryStream and take care of encoding himself.

> Roman

Nickolay