Hi, Bob,

Please see inline further explanations, and answers to your questions.

 
2012-12-12 01:18, Bob Hinden <[email protected]>:

> Rémi,
> 
> Like Brian, I am also trying to understand this.  
> 
> My initial personal conclusion to answer Suresh's question, I don't think 
> this is compatible with the IPv6 address architecture.  
> 
> It proposes to change how interface IDs are constructed (the use of the "g" 
> and "i" bits in the modified EUI-64) and sets up a new IANA registry (it's 
> not clear to me what else would need to be registered).  

Note that it is just a backward-compatible extension, not a change that would 
imply on anything existing today.
 It takes advantage of an unused combination of bits to introduce a specific 
way "to allow development of future technology that can take advantage of 
interface identifiers with universal scope", as envisaged in RFC 4291 itself.
 Appendix A of RFC 4291 "provides examples on the creation of Modified EUI-64 
format-based interface identifiers". The two of them that apply to links having 
universal scope IIDs (i.e. having u = 1) are based on IEEE EUI-64 and IEEE 
EUI-48 at link layer. Both have g=0 (unicast addresses) so that u=g=1 remains 
as an escape pattern for new universal-scope formats. 
 Advantage of this escape pattern over using a reserved OUI in the IEEE EUI-64 
format, a technique mentioned by Brian in his las t mil, is that it leaves 62 
free bits instead of 40. As 4rd needs 48 bits for its own use (an IPv4 address 
and a 16-bit checksum neutrality preserver), it is the first design that needs 
the escape pattern. It leaves plenty of space for other uses of this pattern.
  
 
> This like a very large change for a document that is intended for 
> Experimental status. Making these changes would seem to me more reasonable 
> after the experiment is completed.

 No change of RFC 41 is AFAIK needed:
(a) RFC 5214, and RFC 4214 before it, already specify an IID format that, being 
a backward-compatible use of RFC-4291 format, didn't lead to a modification of 
RFC 4291.
(b) IANA does list in its registries assigned numbers of experimental RFCs.   


> Given this is a tunneling based solution, why not just add a destination 
> header that includes the information you need to make this approach 
> stateless?  Then you wouldn't have to change the way IPv6 interface 
> identifiers are created and avoid any unintended consequenses.

All of this has been carefully reviewed in Softwire, and even more so by the 
six co-authors of widely different origins. No alternative has been found with 
the same desired combination of functional properties.


> Further, I thought we already had solutions to providing IPv4 connectivity 
> over IPv6 (there are some large operators doing this today).  The rhetorical 
> question is why do we need more?

The need for a _stateless_ solution to assign _public_ IPv4 addresses across 
IPv6-only domains, including shared IPv4 addresses, has been abundantly 
discussed and approved in Softwire.
 While MAP-E is now the path chosen for standard track, the WG has found enough 
complementary functional properties in two other solutions to proceed with them 
on experimental track, 4rd and MAP-T.  


Hoping to have given the expected clarifications and answers,
Best regards,
RD



> 
> Thanks,
> Bob
> 
> 
> On Dec 11, 2012, at 6:35 AM, Rémi Després wrote:
> 
>> Hi Brian,
>> 
>> Thank you for the quick comments on this issue.
>> Let me explain more below.
>> 
>> 2012-12-11 11:55, Brian E Carpenter <[email protected]>:
>> 
>>> Hi Suresh,
>>> 
>>>> The 4rd draft (http://tools.ietf.org/html/draft-ietf-softwire-4rd-04)
>>>> describes a solution for providing IPv4 connectivity over IPv6. The
>>>> draft describes the method for mapping 4rd IPv4 addresses to 4rd IPv6
>>>> Addresses. It uses a 4rd specific mark called the V octet in the first 8
>>>> bits of the Interface Identifier. There were some concerns raised in
>>>> softwire about whether such addresses are actually compatible with the
>>>> IPv6 addressing architecture. Whether this is actually compatible with
>>>> the IPv6 addressing architecture is outside the scope of the softwire
>>>> wg. Hence we would like to hear the 6man wg's perspective on this. I
>>>> would like to request the wg to please go over the NOTE in Section 4.5
>>>> page 18, which explains the issue, and over IANA Considerations Section
>>>> 6, and chime in on whether this is acceptable from a 6man perspective.
>>> 
>>> It certainly makes me nervous. The proposed format misuses the "g" bit,
>>> which is supposed to flag multicast.
>> 
>>> Can it be proved beyond doubt that
>>> such an address will never escape a context in which the V octet will
>>> only be interpreted 4rd-style?
>> 
>> No, but this is AFAIK acceptable:
>> - If a 4rd packet enters, in a 4rd domain, a customer site that has no 4rd 
>> CE will be routed in this site like any other IPv6 packet, based on its IPv6 
>> prefix.
>> - It is guaranteed, however, that no properly configured interface will ever 
>> be reached at this address. Reason is precisely that no properly configured 
>> host may have an interface ID starting with the 4rd exclusive /8 prefix.
>> 
>> This prefix is so far called the V octet, for historical reasons, but could 
>> IMHO be advantageously renamed to be more intuitive (e.g. 4rd IID prefix).
>> 
>>> This is not clear to me from the draft.
>>> 
>>> I would be less nervous if you could also set the OUI bits to a reserved
>>> value for this purpose.
>> 
>> Unfortunately, it proved to be impossible: this wouldn't have left 48 bits 
>> for the IPv4 address and the 16-bit checksum-neutrality preserver. Both have 
>> to fit in the IID.
>> 
>>>        For this, the V octet has its "u" and "g" bits of
>>>        [RFC4291] both set to 1, so that they differ from "u" =
>>>        0, reserved for Interface IDs that have local-scope, and
>>>        also differs from "u" = 1 and "g"= 0, reserved for
>>>        unicast Interface IDs using the EUI-64 format.  Bits
>>> 
>>> That should be "modified EUI-64 format"
>> 
>> The proposed 4rd address format introduces, for unicast IPv6 addresses, a 
>> new variant of the "modified EUI-64" IID format 
>> It comes in addition to variants so far defined, namely:
>> - One with u = 0, for the IID privacy option (RFC 4941). 
>> - One with u = 1, for universal IIDs that embed IEEE OUIs and have g = 0 or 
>> unicast addresses (Appendix A of RFC 4291).  
>> 
>> This proposal is expected to be fully in line with the openness expressed in 
>> RFC 4291 by saying "the use of the universal/local bit in the Modified 
>> EUI-64 format identifier is to allow development of future technology that 
>> can take advantage of interface identifiers with universal scope".
>> While reserved OUIs is one way to use this openness, reserved IID prefixes 
>> having u=g=1 is another. 
>> 
>> 
>>> 
>>>        other than "u" and "g", are proposed to be 0, giving V =
>>>        0x03. 4rd is thus the first "future technology that can
>>>        take advantage of interface identifiers with universal
>>>        scope" [RFC4291].
>>> 
>>> I find that sentence hard to swallow. It's recycling the u and g bits
>>> but I don't see how it takes advantage of universal scope. I would
>>> just delete the sentence.
>> 
>> OK.
>> Deleting the sentence in the draft shouldn't be a problem at all. 
>> It was intended as a clarification for 6man, but doesn't add any technical 
>> content.
>> 
>>> 
>>>        As such, it needs to be endorsed by
>>>        the 6man working group and IANA (Section 6).
>>> 
>>> This sentence too can be deleted before RFC publication.
>> 
>> OK.
>> 
>>> 
>>> o  An IPv6 Interface-ID type reserved for 4rd (the V octet of
>>>    Section 4.5).  For this creation of new registry is suggested for
>>>    Interface-ID types of unicast addresses that have neither local
>>>    scope nor the universal scope of Modified EUI-64 format
>>>    [RFC4291]).  This registry is intended to be used for new
>>>    Interface-ID types that may be useful in the future.
>>> 
>>> I don't get this. The possible ug combinations are all theoretically
>>> valid today, aren't they? 11 means globally unique multicast. The only
>>> way you can protect them as a 4rd flag is by using a dummy OUI value,
>>> as far as I can see. I think that is what you should ask for (from
>>> IEEE, not IANA).
>> 
>> As explained above, "globally unique multicast" isn't suitable for a 
>> globally unique _unicast_ address.
>> 
>> If I may add a detail: the proposed new IANA registry would contain IID 
>> prefixes that, in the context of IETF modified EUI-64,  have universal scope 
>> and don't embed IEEE OUIs. The first one would be the 4rd mark 0300::/8. 
>> Others to come may have different lengths (but at least 8 to include u and g 
>> bits)
>> 
>> Best regards,
>> RD
>> 
>> 
>> 
>> 
>>> 
>>> Regards
>>> Brian Carpenter
>>> 
>>> 
>>> --------------------------------------------------------------------
>>> IETF IPv6 working group mailing list
>>> [email protected]
>>> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
>>> --------------------------------------------------------------------
>> --------------------------------------------------------------------
>> IETF IPv6 working group mailing list
>> [email protected]
>> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
>> --------------------------------------------------------------------
> 
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to