Jinmei, I believe your proposed new text at the bottom is correct.
2462bis should not open the door to conflict in future link-layer
specs.

   Brian


JINMEI Tatuya wrote:
On Tue, 18 May 2004 17:46:15 +0200, Brian E Carpenter <[EMAIL PROTECTED]> said:


In this message, I pointed out that there might be possible conflict
between the IPv6 address architecture and link specific documents
(e.g. IPv6 over ethernet), and asked how we can clarify the point.

Unfortunately, I've not seen a direct answer to the question...I now
think it's time to decide how we should deal with this issue in
rfc2462bis.


First of all let's not forget that the addressing architecture is 100%
unambiguous in saying that unicast interface IDs are required to be
64 bits, so a link-specific document that attempted to define a different
ID length would be in direct conflict with RFC 3513. So the question
*cannot* arise unless RFC 3513 is updated or the IESG approves something
that is broken.


I don't think I forgot the point.  I know RFC3513 and RFC 3587 are
100% unambiguous.  I know link-specific documents (e.g. IPv6 over
ethernet) are also 100% unambiguous.  The problem is that the two sets
of unambiguous documents may conflict with each other or at least the
two sets may seem to conflict for implementors.


But at the same time, I guess it is irresponsible to leave the RFC2462
text on this point as is, because we know this issue has been annoying
implementors.


In my view, only because they failed to review the addressing architecture,
which is unambiguous.


I don't think so, because RFC2462 uses link-specific documents as the
primary source about (the length of the) the IFID:

   interface identifier - a link-dependent identifier for an interface
        that is (at least) unique per link [ADDR-ARCH].
[...]
        The exact length of an interface identifier and the way
        it is created is defined in a separate link-type specific
        document that covers issues related to the transmission of IP
        over a particular link type (e.g., [IPv6-ETHER]).

BTW: according to what you were saying, it seems to me that

1. the primary source should actually be the ADDR-ARCH document, not
   link-specific documents
2. link-specific documents must be consistent with ADDR-ARCH.  If not,
   the IESG will reject the document.

If this is the case, I'm just happy with it (is there any direct
reference which states that ADDR-ARCH is the primary?).  Then it will
probably be enough to revise the definition to:

   interface identifier - a link-dependent identifier for an interface
        that is (at least) unique per link [ADDR-ARCH].
[...]
        The exact length of an interface identifier is defined in
        [ADDR-ARCH] based on the address format.  The way it is
        created is defined in a separate link-type specific document
        that covers issues related to the transmission of IP over a
        particular link type (e.g., [IPv6-ETHER]).

perhaps with a note like this:

        The link-type specific document may also define the length of
        the interface identifier.  However, it must be consistent with
        the definition in [ADDR-ARCH].

(If there is a reference that I asked above, it's also good to add the
reference here.)

                                        JINMEI, Tatuya
                                        Communication Platform Lab.
                                        Corporate R&D Center, Toshiba Corp.
                                        [EMAIL PROTECTED]


-------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------

Reply via email to