--- linux-2.3.46/Documentation/CodingStyle	Thu Feb 17 20:45:35 2000
+++ linux-2.3.46/Documentation/CodingStyle.new	Thu Feb 17 20:43:07 2000
@@ -233,3 +233,26 @@
 stable. All options that are known to trash data (experimental write-
 support for file-systems, for instance) should be denoted (DANGEROUS), other
 Experimental options should be denoted (EXPERIMENTAL).
+
+
+		Chapter 8: Struct Initializations
+
+Struct initializations should use named syntax.  It makes for cleaner,
+more compact, self-documenting and more readable code and allows later
+reordering of struct members (for example, to help cache behavior).
+It also allows new struct members to be introduced where they make the
+most sense (say, in the same cache line as other members they are likely
+to be access with).
+
+Since the official kernel compiler is gcc 2.7.2.3, that syntax is
+currently the preferred method.  It is worth noting that C99 syntax
+(.field = initializer,) is also accepted by this compiler.
+
+struct struct_type foo = {
+	field1:		initializer1,
+	field2:		initializer2,
+	field3:		initializer3,
+}
+
+The trailing comma on the last line allows for a homogenous look and easy
+rearrangement (eg for later blocking of some members in an #ifdef).