Now we are using //TRANSLIT option (if supported) for charset conversions. For messages like this one http://trac.roundcube.net/ticket/1486552 This is not a good solution. Some messages (this one too) contains unconvertable sequences. If conversion fails message is displayed as is. For some charsets this is not acceptable. Thunderbird, for example, displays this message properly decoded.

So, question is: should we use //IGNORE instead of //TRANSLIT? Maybe for non-latin charsets only?

ps. //TRANSLIT//IGNORE doesn't work.