Clarified that language codes must be well-formed to match, #8720

Author: svgeesus

selectors-4/Overview.bs

@@ -2640,9 +2640,27 @@ The Language Pseudo-class: '':lang()''</h3>
 	prior to the <i>extended filtering</i> operation.
 	The matching is performed [=ASCII case-insensitively=].
 
-	The <a>language range</a> does not need to be a valid language code to
+	The <a>language range</a> 
+	must be <em>well-formed</em> according to the syntax of BCP 47 in order to match.
+	However, it does not need to be a <em>valid</em> language code to
 	perform this comparison.
 
+	<div class="example">
+		For example,
+		<pre class="lang-css">
+			:lang(åå)
+		</pre>
+		would not match,
+		because it containins non-ASCII characters so is ill-formed.
+
+		On the other hand,
+		<pre class="lang-css">
+			:lang(qq)
+		</pre>
+		could match,
+		even though qq is not a registered language code.
+	</div>
+
 	For this purpose, a wildcard [=language range=] (<code>"*"</code>)
 	does not match elements whose language is not tagged (e.g. <code>lang=""</code>),
 	but does match elements whose language is tagged as undetermined (<code>lang=und</code>).