There is the Encode suite of modules. If you have an unknown encoding, Encode::Guess tries its best to guess the encoding.

XML without an explicit encoding declaration must be UTF-8, but depending on your interpretation of the spec, UTF-16 and UTF-32 are allowed if you use a BOM.

If your encoding is not UTF-8|16|32, you're out of luck. There is NO reliable way to guess charset.

The problem with guessing character encoding is pretty impossible to solve accurately. It's easy to tell the difference between any of the /UTF-\d\d?/ encodings and ASCII, and it's still pretty easy to tell the difference between ASCII and an 8-bit character set or 8-bit character set and /UTF-\d\d?/. Problems begin when you have 8-bit encoded data and you try to guess which one of the dozens of 8-bit encodings it is.

For example, the ISO-8859 series contains no less than fifteen different encodings, plus all old IBM/Microsoft code pages (CP850, CP437; remember DOS?), encodings used by Windows (such as CP-1252), and so on, not to even mention Asian and East European encodings! In a word: good luck.

(I think Shift JIS is also easy to differentiate from ASCII and UTF, but I have no experience with it. It's 8-bit, but uses two bytes for non-ASCII characters.)