[arin-ppml] IPv4 SWIP requirements (?)

[David R Huberman]

Albert,

First, I wanted to say that both as a member of the community (as a 
network operator) and as an AC member, I was exceedingly happy when you 
proposed this draft policy.  I don't agree with everything you have 
written, but I agree with a lot of it, and I think the draft policy 
language is a good starting point.  Just as (or possibly more) 
importantly, I wanted us to have the discussion we're having about the 
purpose of SWIPs. Because as I wrote to rfg, I see the SWIP policy as an 
anachronism, and I think it's time we really dig into the policies and 
figure out where they should go.

Second, I'd like to ask you to please stop referring to specific companies 
by name. We purposely talk in broad concepts because policy applies to 
large swaths of companies -- entire classes of networks -- not just "ISP 
X".  It's fine to say "I have watched a certain cableco not play by the 
rules" to get your point across, for example :)

Now, some specific feedback:

> I wrote this proposal with all seriousness, and do not see it as head of 
> a pin dancing.  When the rules for v4 likely affect less than 5-10% of 
> the total customer base at an ISP, but adding IPv6 elevates it to 100%, 
> this is wrong, and this does deserve a serious shot at repair.  I would 
> like to see the percentages of v6 customers subject to this rule to be 
> roughly the same as the current number of v4 customers subject to the 
> rule.

I agree.

[snip of the logical argument backing up the claims of inequity between v4 
and v6 policy.]

> How can you expect such a rule to be followed, and is it reasonable to 
> subject the majority of the access customers to this rule for v6, when 
> it has NEVER been the rule in v4? I have never seen anyone propose SWIP 
> at the /32 level.  The current v6 standard of 100% SWIP is UNREASONABLE. 
> This is why I am proposing a change in the standard.

Again, I agree.

> Reasonable rules are more likely to be complied with, and whatever the 
> rule is, I agree that the rules should not be ignored, and also agree 
> that in fact, it is widely ignored.  If it were made more reasonable, I 
> have hope that it might also be followed more.
>
> If the Registration rule was made closer to the current v4 rule, such that 
> does not catch most access provider customers, there will be fewer addresses 
> to SWIP, and I believe it will be more likely than the current rule to be 
> followed, as the number of assignments requiring registration will be vastly 
> decreased from the current standard of 100% of v6.

That makes sense. I'm not sure I agree (and I'm not sure I disagree), but 
I can accept your posulation.

> In answer to the question as to the purpose of this proposal, it is to make 
> the rules for SWIP more equal between IPv4 and IPv6.  Currently, IPv4 only 
> requires SWIP for a /29 or more, leaving the majority of access circuits 
> without any SWIP requirement whatsoever.  This is NOT currently true for 
> IPv6, which the policy manual requires registration for a /64 or more of 
> space, which is basically 100%.

So to make them more equal, I actually preferred your approach - lower the 
v4 threshold and lower the v6 threshold. I think anything smaller than a 
/24 is not helpful.  But I suspect I'm in a small minority there, so I'll 
just assume it's never going to happen :)

For this policy, I'd prefer a SWIP if it's a /48 or larger.  Subnets 
smaller than a /48 are not typically going to ever be independently 
routed, and that's my interest in SWIP policy.

Thanks again for submitting and defending your policy proposal!

David