text/css
During the breaks of the meetings and via IRC, Henri, Simon, and I, with a little help from Travis to test Internet Explorer, figured out the new set of rules to decode a text/css
resource taking into account the lessons learned from the Encoding Standard. The result is now documented in CSS Syntax Module: The input byte stream thanks to Tab Atkins.
Relative to CSS 2.1 it now gives priority to the byte-order-mark, as is also done in HTML and JavaScript, no longer allows sniffing utf-16 out of an ASCII-encoded @charset
rule, and no longer supports various obsolete concepts, such as utf-32.