Currently, Decoder::decode() only looks for "@charset" bytes when decoding CSS - it should also support UTF-16, UTF-32, and, theoretically, even EBCDIC. See bug URL for the specification.
Created attachment 10441 [details] test case