Incorrect behavior of truncate on utf-8 strings

I have a problem with truncate on Russian strings. For example if s is "Словарь" then

s = truncate(product.description, :length => 9)

returns correct string with bytes
208 161 208 187 208 190 46 46 46

But if I truncate to 10 characters, it returns the following byte sequence
208 161 208 187 208 190 208 46 46 46
where 208 before last three dots is a start of next UTF-8 character. So it looks like truncate cuts based on byte length and rounds up and gets wrong bytes.

This results in 'invalid byte sequence in UTF-8' exception.

I'm using Rails 2.3.5 and Ruby 1.9.1 on Ubuntu 9.10.