U+0080 as non-ASCII for CSS tokenization

Attached is a test case showing that some browsers consider U+0080 
“non-ASCII” for the purpose of CSS tokenization, while the CSS spec say 
that non-ASCII is U+00A0 and above.

My proposal is to change the CSS definition of non-ASCII to “U+0080 and 
above” to match implementations (as well as any definition of ASCII 
which stops at most at 0x7F.)

Discussion starts here:
http://lists.w3.org/Archives/Public/www-style/2013Jan/0394.html

-- 
Simon Sapin