RE: [dhcwg] FW: RFC3315 question

["Bernie Volz \(volz\)" <[e-mail]>]

Ah, good that this is finally being cleaned up (I recall there was some
past discussion on this). Reference to RFC 826 was very poor and really
should have referenced the IANA pages (the Number Hardware Type (hrd)
section of http://www.iana.org/assignments/arp-parameters).

Perhaps a reference to http://www.ietf.org/rfc/rfc2469.txt might also
help to address the canonical address issue? At least the describes
canonical addresses.

Link layer also confused one major vendor into thinking they should use
the EUI-64 identifier, so perhaps an example wouldn't hurt as well (we
have an example for DUID-EN, so why not for the others?).

- Bernie 

-----Original Message-----
From: Ralph Droms (rdroms) 
Sent: Friday, February 16, 2007 4:53 PM
To: dhcwg
Subject: [dhcwg] FW: RFC3315 question

Alain Durand asked me a couple of questions about the definition of the
DUID
formats based on link-layer addresses in RFC 3315.  The first question
is
about this text, which appears in both section 9.2, DUID Based on
Link-layer
Address Plus Time [DUID-LLT], and section 9.4, DUID Based on Link-layer
Address [DUID-LL]:

   The link-layer
   address is stored in canonical form, as described in RFC 2464.

Alain checked RFC 2464, and he can't find how a link layer address is
supposed to be stored in canonical form. The closest thing he found is:
 
   The Source/Target Link-layer
   Address option has the following form when the link layer is
   Ethernet.

                    0                   1
                    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5
                   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
                   |     Type      |    Length     |
                   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
                   |                               |
                   +-          Ethernet           -+
                   |                               |
                   +-           Address           -+
                   |                               |
                   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

And, RFC 2464 only describes Ethernet and is silent on other link-layer
addresses.

There is also an issue with the text:

   The hardware type MUST be a
   valid hardware type assigned by the IANA, as described in RFC 826.

...but, RFC 826 makes no reference to IANA.

Ted Lemon and I exchanged e-mail about the DUID issue, and I've included
Ted's response below.

Regarding the IANA/RFC 826 issue, I think it can be addressed by an
errata:

OLD:

   The hardware type MUST be a valid hardware type
   assigned by the IANA as described in RFC 826 [14].

NEW:

   The hardware type MUST be a valid hardware type from the list of
   ARP hardware types maintained by IANA at <insert URL here>

- Ralph

=====

To be honest, I have no idea where the RFC2464 reference came from.
I'd never read that RFC, and wasn't even aware of its existence.
The word 'canonical' is only used twice in the RFC, and nowhere in
the RFC is the meaning of 'canonical' described.   The point Alain
raises is a really good one, and applies to RFC2464 as well, to some
extent - it should define 'canonical'.

However, RFC2464 does give us some clue about this in section 2.5.1,
Interface Identifiers, on page 8, where it describes in detail how to
construct an interface identifier.   This does tell us what the
author of RFC2464 thought the canonical form is.  But Alain's still
right that the reference to RFC2464 is a mistake in isolation,
because it leaves out all other possible frame header formats.
Having said that, the problem remains - without specifying the
arrangement of bits, you are left guessing.   People seem to have
muddled through with RFC2131/2132, and I think they probably will
with 3315 as well, but it could have been better specified.

I think the correct fix for the RFC2464 problem is to add some text
like what's in RFC2373:

    The details of forming interface identifiers are defined in the
    appropriate "IPv6 over <link>" specification such as "IPv6 over
    Ethernet" [ETHER], "IPv6 over FDDI" [FDDI], etc.

We'd change it to this:

    The details of how the link-layer address is formatted are
defined in the
    appropriate "IPv6 over <link>" specification such as "IPv6 over
    Ethernet" [RFC2464], "IPv6 over FDDI" [RFCnnnn], etc.
Specifically, the
    link-layer address should appear in the DUID-LLT option just as
it does
    in the frame header for that particular link type.  For example the
    link-layer address for an ethernet interface would be formatted
in the
    same way as the Destination Ethernet Address field of the
ethernet frame
    as described in the section of "IPv6 over Ethernet" [RFC2464]
titled "Frame
    Format."

This doesn't really solve the problem, because RFC2464 isn't specific
about it either, but it gets rid of the erroneous reference to
RFC2464 as the _only_ place to go, and it also gives the implementor
a valuable clue for how to format the DUID-LLT - look at a frame you
captured live on the wire.   That way.   That's lame, but it's better
than nothing.   Really, RFC2464 should be more particular about
referencing the document (presumably IEEE) that defines the 802.11
header format.   You can see that the author had access to that
document by reading section 2.5.1, but unfortunately s/he never
references it.

_______________________________________________
dhcwg mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/dhcwg

_______________________________________________
dhcwg mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/dhcwg