1. We should be able to do `diff` even for source and destination having different encodings and/or charsets.
Add option for source encoding and destination encoding. 2. The conversion to common encoding should be even with the option to not fail on wrong or unconvertible characters (like //IGNORE in GNU iconv), that is replacing unknown characters with a placeholder character. This is useful to compare wrongly encoded files. 3. More generally we can add filter for every compared file first to pass through the filter. Item 1 can be implemented by passing `iconv` or `recode` command as such a filter.