Re: BOM requirement in UTF-16

>If a BOM appears, it determines the encoding.

According to which standard?  Unicode says (section 13.6):

  Where the character set information is explicitly marked, such as in
  UTF-16BE or UTF-16LE, then all U+FEFF characters, even at the very
  beginning of text, are to be interpreted as zero width no-break
  spaces.

>XML's whitespace vocabulary is very limited.  Such a character is not
>allowed in an XML document, so the document would not be well-formed.

You're right, it would not be allowed at the start of a document
because it is not an XML whitespace character.  (It is allowed in text
content however.)

-- Richard