[arin-ppml] Draft Policy ARIN-2017-5: Equalization of Assignment Registration requirements between IPv4 and IPv6 - updated 2017-07-21
Owen DeLong
owen at delong.com
Mon Jul 24 14:45:47 EDT 2017
> On Jul 24, 2017, at 04:03 , hostmaster at uneedus.com wrote:
>
> /47 or more addresses is intended to be /47, /46 ..... /1 and not the reverse. The current language is "/64 or more", and I read that same phrase as /64, /63 ..... /1. For comparison, the current IPv4 language is "/29 or more", and that seems clear to mean /29, /28 ..... /1, and not the reverse. I do not think the "/X or more" language in the current NRPM or the proposal is unclear.
>
> As far as the remainder, I read this draft as two statements connected with an OR. Thus there are 2 classes of assignments that are required to be SWIP'ed. They are:
>
> 1) ANY IPv6 assignment containing a /47 or more addresses.
> 2) ANY sub-delegation of any size that will be individually announced.
>
> Due to rules for the GRT, this means /49 or less is not covered by 2). However, should the rules for the GRT ever change, 2) will cover it. The intent of 2) is to require SWIP on ALL sub-delegations, without expressing that level, which is outside of ARIN control.
There are no rules for the GRT. There are only each ASNs rules for what they will accept and announce and what ends up in any particular routing table is the intersection of routes accepted and routes advertised by peers.
Just a quick sampling from Route-views:
* 2001:218:200e:100::/56
2a01:73e0::1 0 47872 2914 i
* 2001:67c:22dc:def1::1
0 31019 41887 2914 i
* 2c0f:fc00::2 0 3741 2914 i
* 2001:d98::19 0 18106 4657 2914 i
* 2001:418:0:1000::f000
17346 0 2914 i
* 2a00:1728::2d:bbbb
0 0 34224 2914 i
*> 2001:418:0:1000::f002
11266 0 2914 i
* 2001:218:200e:200::/56
2a01:73e0::1 0 47872 2914 i
* 2001:67c:22dc:def1::1
0 31019 41887 2914 i
* 2c0f:fc00::2 0 3741 2914 i
* 2001:d98::19 0 18106 4657 2914 i
* 2001:418:0:1000::f000
17346 0 2914 i
* 2a00:1728::2d:bbbb
0 0 34224 2914 i
*> 2001:418:0:1000::f002
11266 0 2914 i
* 2001:218:200e:300::/56
2a01:73e0::1 0 47872 2914 i
* 2001:67c:22dc:def1::1
0 31019 41887 2914 i
* 2c0f:fc00::2 0 3741 2914 i
* 2001:d98::19 0 18106 4657 2914 i
* 2001:418:0:1000::f000
17346 0 2914 i
* 2a00:1728::2d:bbbb
0 0 34224 2914 i
*> 2001:418:0:1000::f002
11266 0 2914 i
* 2001:218:200e:400::/56
2a01:73e0::1 0 47872 2914 i
* 2001:67c:22dc:def1::1
0 31019 41887 2914 i
* 2c0f:fc00::2 0 3741 2914 i
* 2001:d98::19 0 18106 4657 2914 i
* 2001:418:0:1000::f000
17346 0 2914 i
* 2a00:1728::2d:bbbb
0 0 34224 2914 i
*> 2001:418:0:1000::f002
11266 0 2914 i
So clearly it is possible to announce at least a /56 and even maybe a /126:
*> 2001:218:4000:5000::338/126
2001:d98::19 0 18106 132602 ?
*> 2001:218:4000:5000::368/126
2001:d98::19 0 18106 132602 ?
*> 2001:218:4000:5000::42c/126
2001:d98::19 0 18106 132602 ?
(though that doesn’t seem to have gotten very far)
But this /64 got a bit further:
* 2001:254:1:b::/64
2001:d98::19 0 18106 4635 10103 3662 3662 3662 3662 3662 3662 3662 ?
* 2402:c100::104 0 23673 4635 10103 3662 3662 3662 3662 3662 3662 3662 ?
*> 2404:cc00:1::1 0 24441 4635 10103 3662 3662 3662 3662 3662 3662 3662 ?
As did this one:
* 2001:2b8:0:ffd0::/64
2001:b08:2:280::4:100
0 3277 3267 2603 17579 1237 17832 i
* 2001:d98::19 0 18106 4635 17579 1237 17832 i
* 2402:c100::104 0 23673 4635 17579 1237 17832 i
*> 2404:cc00:1::1 0 24441 4635 17579 1237 17832 i
I wouldn’t say that announcing longer prefixes is reliable, but it’s quite clear to me that people are doing so with at least some degree of success getting them propagated beyond their local ASN or even their direct peers.
> I do agree that "will be" is a problem as to WHEN to SWIP. This is best corrected by changing that to "is". 6.5.5.2 the next section makes it clear that the time frame to SWIP is within 7 calendar days of the assignment being made.
I think “will be” is fine as I believe it conveys an obligation to SWIP if the delegation is intended to be announced separately when it is delegated vs. no requirement to SWIP if until such an intent exists.
> As far as John's comment, this proposal began with a suggestion that changed the v4 requirement as well, making both "more than 16" networks or IPv4 addresses. Since changing the v4 language from 8 addresses to more than 16 addresses was clearly not desired by the community, the v4 language was removed from the draft. The comments still reflect that the equality of the policy between v4 and v6 was the original idea. You are correct that this draft now is only about changing the v6 part. Are you suggesting that the older portions of description that are no longer in it, due to the community input needs to be removed?
John’s comments are still correct that any theory of equality between the two protocols is a false equivalence which cannot, in practice, be achieved. IPv6 policy can certainly be improved and made no more onerous than IPv4 policy, but any argument of equivalence is doomed to failure due to the staggering differences in the address management strategies and subnetting practices of the two protocols.
Owen
>
> Albert Erdmann
> Network Administrator
> Paradise On Line Inc.
>
> On Sun, 23 Jul 2017, Scott Leibrand wrote:
>
>> No. It says:
>>
>> Each static IPv6 assignment containing a /47 or more addresses, or sub-delegation of any size that will be individually announced, shall be registered in the WHOIS directory
>>
>> I read that as:
>>
>> Each static IPv6 assignment containing a /47 or more addresses, and each sub-delegation of any size that will be individually announced, shall be registered in the WHOIS directory
>>
>> So a /48 sub-delegation shall be registered if it will be individually announced.
>>
>> You said:
>>> To be explicit, to me, "/47 or more addresses, or sub-delegation of any size that will be individually announced," refers to /47s, /46s, /45s ... and not /48s, /49s, /50s, etc.
>>
>>
>> That entire quoted clause refers to /48s (or longer, if such becomes possible) that will be individually announced. So your clarification, as originally stated, does not match my reading of the text. However, it now sounds like that's not what you meant, and you were trying to say:
>>
>>> To be explicit, to me, "/47 or more addresses," refers to /47s, /46s, /45s ... and not /48s, /49s, /50s, etc.
>>
>> If that's what you actually meant, then yes, that is the way to read "more": as equivalent to "/47 or shorter".
>>
>>
>> Leif, it seems like we have some potential ambiguity in the new text as to whether "or sub-delegation of any size" is part of the subject of the sentence, or a subordinate clause to "containing". Does my rewrite above clarify your intent? Or were you intending it to mean "Each static IPv6 assignment containing a ... sub-delegation of any size that will be individually announced, shall be registered in the WHOIS directory"? That interpretation makes no sense to me, as the sub-delegation clause would be redundant. So if that's what you meant, I'd appreciate some clarity on what it is intended to accomplish.
>>
>> Scott
>>
>>> On Jul 23, 2017, at 8:35 PM, John Springer <3johnl at gmail.com> wrote:
>>>
>>> Thanks, Scott,
>>>
>>> Are we energetically agreeing? You scared me there for a second. /48s are excluded, unless they are part of a "subdelegation of any size that will be individually announced". Yes.
>>>
>>> How is that defined by the way? Will be individually announced in 2 years, 2 days, right now?
>>>
>>> On another matter, this problem statement has been making me uneasy all along, but because it was only required to be clear and in scope to be accepted as Draft Policy, it was not appropriate for me to object. This seems like as good a time as any to address some concerns, which are my opinions only.
>>>
>>> "Current ARIN policy has different WHOIS directory registration requirements for IPv4 vs IPv6 address assignments."
>>>
>>> This is a correct statement! It is not a problem however, nor is it sufficient motive for trying to solve a problem, per se.
>>>
>>> "IPv4 registration is triggered for an assignment of any address block equal to or greater than a /29 (i.e., eight IPv4 addresses).
>>> In the case of IPv6, registration occurs for an assignment of any block equal to or greater than a /64,"
>>>
>>> Two facts. The second is undoubtedly a great pity, but to entangle these logically is a fallacy of inconsistency, specifically a false equivalence. These two facts are unrelated. It does not help the case to try to make them interdependent. And it is not needed. All that is being attempted is to modify V6 SWIP requirements. Do that. And DO NOT settle on 8 subnets.
>>>
>>> "which constitutes one entire IPv6 subnet and is the minimum block size for an allocation."
>>>
>>> I think I'm looking at current text. How did this make it this far? One ENTIRE IPV6 subnet? There are lots of entire V6 subnets all the way from /0 to /128. What does that have to do with anything? And, yeah, the SWIP boundary being the so called "minimum" allocation seems broken, but that is its own thing.
>>>
>>> "There is no technical or policy rationale for the disparity, which could serve as a deterrent to more rapid IPv6 adoption."
>>>
>>> Possibly true, but irrelevant. There is no technical or policy rationale for them being alike either, nor is there any reason to suppose that if they were, folks would adopt V6 faster. SWIPing /64 is definitely wrong for V6. Concentrate on that. We can make policy for V6 without needing to refer to V4.
>>>
>>> "The purpose of this proposal is to eliminate the disparity and corresponding adverse consequences."
>>>
>>> With respect, it is not. The disparity does not qualify as a logical motive. The brokenness of SWIPing /64s does not require injustice and if /64 SWIPing is a deterrent to V6 adoption, that is its own good and sufficient reason. If you had to refer to an analogy, you could say, "SWIPing /64s is analogous to SWIPing /32s and that seems dumb".
>>>
>>> So all you need is:
>>>
>>> Problem Statement: SWIPing IPV6 /64s is the problem. The purpose of this proposal is to pick a different number.
>>>
>>> Policy statement:
>>> 1) Alter section 6.5.5.1 "Reassignment information" of the NRPM to strike "/64 or more addresses" and change to "/47 or more addresses, or subdelegation of any size that will be individually announced," and
>>> 2) Alter section 6.5.5.3.1. "Residential Customer Privacy" of the NRPM by deleting the phrase "holding /64 and larger blocks"
>>>
>>> BUT. These observations do not appear to have any effect one way or the other on the policy text. To me, picking a different number does not have anything to do with disparity, but so what? Changing the IPV6 SWIP threshold is not unfair and partial if someone makes unfounded assertions regarding linkages between v4 and V6. And it is not technically unsound to make fallacious observations if they are kind of orthogonal to the meat of the matter.
>>>
>>> So, still support. I'd rather see it simpler, but I guess I can tolerate a little hand waving.
>>>
>>> Writing solely on my own behalf,
>>>
>>> John Springer
>>>
>>>> On Fri, Jul 21, 2017 at 9:15 PM, Scott Leibrand <scottleibrand at gmail.com> wrote:
>>>>> On Jul 21, 2017, at 8:31 PM, John Springer <3johnl at gmail.com> wrote:
>>>>>
>>>>> I support this Draft Policy as re-written.
>>>>>
>>>>> I shared the author's distaste for the requirement that IPV6 /64s be SWIP'd, but was not reassured when the discussion veered to consider prefixes between /48 and /64. AFAIK, ISPs have long been encouraged to apply for their allocations based on the idea of assigning a /48 to each 'customer' to provide room for unknown technologies, among other things. I did not wish to endanger that premise, but current language appears to moot that concern.
>>>>>
>>>>> To be explicit, to me, "/47 or more addresses, or sub-delegation of any size that will be individually announced," refers to /47s, /46s, /45s ... and not /48s, /49s, /50s, etc.
>>>>
>>>> That's not what it says. It says /48s (or longer) should be individually SWIPped if they're going to be announced. Otherwise there's no reason for the extra clause.
>>>>
>>>> Blocks in the GRT need to be SWIPped to the announcing party if that's a different organization from the holder of the larger block.
>>>>
>>>> -Scott
>>>>
>>>>>
>>>>>> On Fri, Jul 21, 2017 at 9:44 AM, Leif Sawyer <lsawyer at gci.com> wrote: Happy Friday, everybody.
>>>>>>
>>>>>>
>>>>>>
>>>>>> As promised, here is the latest rewrite of the draft policy below, and it will soon be updated at:
>>>>>>
>>>>>> https://www.arin.net/policy/proposals/2017_5.html
>>>>>>
>>>>>>
>>>>>>
>>>>>> There are two changes noted in the policy statement: the first of which reflects what seems to be the current
>>>>>>
>>>>>> consensus of the PPML regarding netblock sizing; the second is to strike language that may be read as either restrictive
>>>>>>
>>>>>> or non-operational.
>>>>>>
>>>>>>
>>>>>>
>>>>>> ----
>>>>>>
>>>>>>
>>>>>>
>>>>>> Problem Statement:
>>>>>>
>>>>>> Current ARIN policy has different WHOIS directory registration requirements for IPv4 vs IPv6 address assignments.
>>>>>>
>>>>>> IPv4 registration is triggered for an assignment of any address block equal to or greater than a /29 (i.e., eight IPv4 addresses).
>>>>>>
>>>>>> In the case of IPv6, registration occurs for an assignment of any block equal to or greater than a /64, which constitutes one entire IPv6 subnet and is the minimum block size for an allocation.
>>>>>>
>>>>>> Accordingly, there is a significant disparity between IPv4 and IPv6 WHOIS registration thresholds in the case of assignments, resulting in more work in the case of IPv6 than is the case for IPv4.
>>>>>>
>>>>>> There is no technical or policy rationale for the disparity, which could serve as a deterrent to more rapid IPv6 adoption.
>>>>>>
>>>>>> The purpose of this proposal is to eliminate the disparity and corresponding adverse consequences.
>>>>>>
>>>>>>
>>>>>>
>>>>>> Policy statement:
>>>>>>
>>>>>> 1) Alter section 6.5.5.1 "Reassignment information" of the NRPM to strike "/64 or more addresses" and change to "/47 or more addresses, or sub-delegation of any size that will be individually announced,"
>>>>>>
>>>>>> and
>>>>>>
>>>>>> 2) Alter section 6.5.5.3.1. "Residential Customer Privacy" of the NRPM by deleting the phrase "holding /64 and larger blocks"
>>>>>>
>>>>>>
>>>>>>
>>>>>> Comments:
>>>>>>
>>>>>> a. Timetable for implementation:
>>>>>>
>>>>>> Policy should be adopted as soon as possible.
>>>>>>
>>>>>>
>>>>>>
>>>>>> b. Anything else:
>>>>>>
>>>>>> Author Comments:
>>>>>>
>>>>>> IPv6 should not be more burdensome than the equivalent IPv4 network size.
>>>>>>
>>>>>> Currently, assignments of /29 or more of IPv4 space (8 addresses) require registration
>>>>>>
>>>>>> The greatest majority of ISP customers who have assignments of IPv4 space are of a single IPv4 address which do not trigger any ARIN registration requirement when using IPv4.
>>>>>>
>>>>>> This is NOT true when these same exact customers use IPv6, as assignments of /64 or more of IPv6 space require registration.
>>>>>>
>>>>>> Beginning with RFC 3177, it has been standard practice to assign a minimum assignment of /64 to every customer end user site, and less is never used.
>>>>>>
>>>>>> This means that ALL IPv6 assignments, including those customers that only use a single IPv4 address must be registered with ARIN if they are given the minimum assignment of /64 of IPv6 space.
>>>>>>
>>>>>> This additional effort may prevent ISP's from giving IPv6 addresses because of the additional expense of registering those addresses with ARIN, which is not required for IPv4.
>>>>>>
>>>>>> The administrative burden of 100% customer registration of IPv6 customers is unreasonable, when such is not required for those customers receiving only IPv4 connections.
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> ---
>>>>>>
>>>>>>
>>>>>>
>>>>>> Leif Sawyer
>>>>>>
>>>>>> Advisory Council
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________ PPML You are receiving this message because you are subscribed to the ARIN Public Policy Mailing List (ARIN-PPML at arin.net). Unsubscribe or manage your mailing list subscription at: http://lists.arin.net/mailman/listinfo/arin-ppml Please contact info at arin.net if you experience any issues.
>>>>>
>>>>> _______________________________________________ PPML You are receiving this message because you are subscribed to the ARIN Public Policy Mailing List (ARIN-PPML at arin.net). Unsubscribe or manage your mailing list subscription at: http://lists.arin.net/mailman/listinfo/arin-ppml Please contact info at arin.net if you experience any issues.
>>>
>>
> _______________________________________________
> PPML
> You are receiving this message because you are subscribed to
> the ARIN Public Policy Mailing List (ARIN-PPML at arin.net).
> Unsubscribe or manage your mailing list subscription at:
> http://lists.arin.net/mailman/listinfo/arin-ppml
> Please contact info at arin.net if you experience any issues.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.arin.net/pipermail/arin-ppml/attachments/20170724/abc629bb/attachment.htm>
More information about the ARIN-PPML
mailing list