Ether types (was: Re: GRE specification going to last call)

[Tim Gleeson <tgleeson@cisco.com>]

At 10:21 00/1/11 -0800, David Meyer wrote:
> 
> > The MSDP WG has the same problem.  We've recently asked IANA for a GRE 
> > Protocol Type value for MSDP.
> 
> 	It seems that this language is problematic. It should be
> 	probably removed. The idea came from comments from the
> 	the original authors that if we do an GREv2, we should
> 	revisit the use of DIX ether types for the protocol field.
> 	
> 	Comments on removing or softening this text:
> 
> "The IANA SHOULD NOT encourage the assignment of additional ETHER
> TYPES (GRE Protocol Types) for use with GRE."

Two issues.

Firstly, why should it matter whether IANA (or anyone else) assigns 
*additional* ether types or not? This should not affect any existing systems.

Secondly, I'm a little confused and/or ignorant about what exactly "DIX 
ether types" are and who is authoritative for them.

RFC2074 (for example), in the ether2 PROTOCOL-IDENTIFIER section says: "DIX 
Ethernet, also called Ethernet-II... The authoritative list of Ether Type 
values is identified by the URL: 
ftp://ftp.isi.edu/in-notes/iana/assaignments/ethernet-numbers."

http://www.cavebear.com/CaveBear/EtherNet/ (last updated 1998/04) (cited by 
"ethernet-numbers") says "The Ethernet Type values are managed by Xerox" 
but also lists "contributions from network managers and vendors" and has a 
reference to the rather ancient RFC1010.

RFC1701 says "In general, the value will be the Ethernet protocol type 
field for the packet. Additional values may be defined in other documents." 
It then gives a list of types, the only ones which conflict with 
"ethernet-numbers" are those below 05DC i.e. SNA, OSI, and PUP.  It also 
says "Future protocol types must be taken from DIX ethernet encoding... See 
the IANA list of Ether Types for the complete list of these values." Is 
this an indirect statement that "ethernet-numbers" is the definitive DIX 
encoding?

The new "gre-update-02" notes "XEROX" and references (the old) RFC1700, 
which does not assign authority, but says you should contact XEROX for an 
ether type.

The "ethernet-numbers" document itself says "The following list of 
EtherTypes is contributed unverified information from various sources." Not 
a very strong assertion of authority. It also says you should contact the 
IEEE Registration Authority if you want an ether type.

http://standards.ieee.org/regauth/ethertype/type-tut.html asserts "the 
value of the type field is obtained from the IEEE Type Field Registrar. New 
values obtained from the IEEE Type Field Registrar will not interfere with 
the existing Type field assignments from Xerox or the IEEE. Former 
assignments are still valid." Whether this is authoritative or not I can't 
say, but it's certainly the most assertive statement I've found on the topic.

But the IEEE don't seem very forthcoming. They state "It is important to 
realize that this is a partial listing of all assigned EtherType Field. Not 
all recipients wish to publish their assignment at this time." The actual 
list of public assignments is at 
http://standards.ieee.org/regauth/ethertype/type-pub.html
But this only lists the organizations to which assignments has been made, 
e.g. "0800-0803" (including 0800 for IPv4) is listed simply as "Xerox". 
Also "86DD" (IPv6) is not listed here at all, but is of course listed in 
"ethernet-numbers". This may be because it was a "secret" assignment, but 
it sounds unlikely since it came from IANA.

Any answers?

As for GRE, can't we just say (words to the effect) "use what was in 
[RFC1701] and what is in [ethernet-numbers]" and remove the section about 
"NOT encourage"?

Tim