[css2] Corrected the BOM errata. (BOM only overrides if the encoding is

bert-github · bert-github · commit 5c4f65950fba · 2013-05-02T14:01:28.000Z
already known to be UTF-based.) And added a note mentioning that a BOM
in UTF-16BE, UTF-16LE, UTF-32BE or UTF-32LE is an error in Unicode.

--HG--
extra : convert_revision : svn%3A73dc7c4b-06e6-40f3-b4f7-9ed1dbc14bfc/trunk%403274
diff --git a/css2/syndata.src b/css2/syndata.src
@@ -1,6 +1,6 @@
 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN">
 <html lang="en">
-<!-- $Id: syndata.src,v 2.193 2013-05-02 13:02:40 bbos Exp $ -->
+<!-- $Id: syndata.src,v 2.194 2013-05-02 14:01:28 bbos Exp $ -->
 <head>
 <title>Syntax and basic data types</title>
 <!--script src="http://www.w3c-test.org/css/harness/annotate.js#CSS21_DEV" type="text/javascript" defer></script-->
@@ -1409,23 +1409,15 @@ encoding::default|default::character encoding">character
 encoding</span> (from highest priority to lowest):
 </p>
 <ol>
-<li><span class="index-inst">BOM</span>
 <li>An HTTP "charset" parameter in a "Content-Type" field
 (or similar parameters in other protocols)</li>
-<li><span
+<li><span class="index-inst">BOM</span> and/or <span
 class="index-inst">@charset</span> (see below)</li>
 <li><code>&lt;link charset=""&gt;</code> or other metadata from the linking mechanism (if any)</li>
 <li>charset of referring style sheet or document (if any)</li>
 <li>Assume UTF-8</li>
 </ol>
 
-<p class=note>Note that it is not possible to use a 1-byte character
-encoding and start the CSS file with the characters 255 and 254 in
-either order, because the two characters will be interpreted as a
-BOM. E.g., "&#255;" and "&#254;" in ISO-8859-1, "&#729;" and "&#355;"
-in ISO-8859-2, etc. Authors should start such files with something
-else, e.g., a space.
-
 <p>Authors using an <span class="index-inst">@charset</span> rule must
 place the rule at the very beginning of the style sheet, preceded by
 no characters. (If a byte order mark is appropriate for the encoding
@@ -1452,6 +1444,27 @@ registry.
 <p>User agents must support at least the <span
 class="index-inst">UTF-8</span> encoding.
 </p>
+
+<p>If rule 1 above (an HTTP "charset" parameter or similar) yields a
+character encoding and it is one of UTF-8, UTF-16, UTF-16BE, UTF-16LE,
+UTF-32, UTF-32BE or UTF-32LE, then a BOM, if any, at the start of the
+file overrides that character encoding, as follows:
+
+<table>
+<thead>
+<th><th>First bytes (hexadecimal)  <th>Resulting encoding
+<tbody>
+<tr><td>00 00 FE FF  <td>UTF-32, big-endian
+<tr><td>FF FE 00 00  <td>UTF-32, little-endian
+<tr><td>FE FF        <td>UTF-16, big-endian
+<tr><td>FF FE        <td>UTF-16, little-endian
+<tr><td>EF BB BF     <td>UTF-8
+</table>
+
+<p class=note>Note that, if rule 1 yields UTF-16BE, UTF-16LE, UTF-32BE
+or UTF-32LE, a BOM at the start of the file is an error. (Unicode
+forbids a BOM in such files).
+
 <p>User agents must ignore any @charset rule not at the beginning of the
 style sheet. When user agents detect the character encoding using the
 BOM and/or the @charset rule, they should follow the following rules:
