[Guy Harris <gharris@xxxxxxxxx>]

On Fri, Nov 08, 2002 at 08:54:52AM +0100, Hannes Gredler wrote:
> On Thu, Nov 07, 2002 at 02:13:05PM -0800, Guy Harris wrote:
> | On Thu, Nov 07, 2002 at 11:01:43AM +0100, Hannes Gredler wrote:
> | > pls find a patch that shows now OSI in GRE tunneled packets
> | > that the payload is OSI and not "unknown";
> | 
> | Checked in (although I got the OSI SAP value from "llcsaps.h" rather
> | than adding it to "greproto.h").
> 
> thats one of the odd things in GRE - the protocol ID of the payload
> is actually an ethertype

...except for the ones that aren't.

To quote RFC 1701:

   The following are currently assigned protocol types for GRE.  Future 
   protocol types must be taken from DIX ethernet encoding.  For
   historical reasons, a number of other values have been used for some 
   protocols.  The following table of values MUST be used to identify
   the following protocols:

       Protocol Family                     PTYPE
       ---------------                     -----
       Reserved                            0000
       SNA                                 0004
       OSI network layer                   00FE

		...

       Raw Frame Relay                     6559

		...

6559 doesn't appear in

	http://standards.ieee.org/regauth/ethertype/type-pub.html

Several items in the Ethertype list at

	http://www.iana.org/assignments/ethernet-numbers

cite "[RFC1701]" as a reference, which seems to indicate that they first
became ethertypes as a result of being GRE protocol types.

Now, perhaps the ones that are in the IANA document have really been
assigned as Ethertypes, so that one could use them to, for example, do
Frame Relay over Ethernet without the Ethertype Gods smiting you, but
00FE isn't a valid Ethertype so it's just a GREism.