RT Encoding Problem

I have stumbled upon an encoding problem in RT. We are using RT in a
multilingual environment, where some of the users will be posting
requests in a non-latin based alphabet (specifically, greek).

Ticket entry and comments generation is fine, in any language, as long
as they are done solely via e-mail. Web-based interaction is
problematic, though.

When a ticket is entered via the web-based form, the subject is always
messed up when in a non-latin based alphabet. The body of the ticket is
OK though—so I guess this must have to do with the db set-up. I
checked that ticket bodies are BLOB/TEXT, whereas subject lines are
VARCHARS. In the RT documentation it is mentioned that “some languages
however do require additional changes to get around an implicit
character-set assumption in the RT 2.0.x series”. Any ideas what changes?

Thanks a lot,