Good questions, Remco! Thanks, Géza On Mon, May 7, 2012 at 7:34 PM, Remco Van Mook <Remco.vanMook@eu.equinix.com

wrote:

<all hats off>

Alright then, for the sake of argument I'll oppose until I see some convincing numbers. Back in the original last /8 discussion the rationale for choosing a /22 was that it would get us about 16k final allocations, or 1 for every NCC member and room for the membership to double in size. Now, we have a number of new realities:

- final /8 applies to all v4 address space when it kicks in, including space that gets returned later; - It also applies to v4 address space that has not been allocated or assigned by RIPE NCC at that date; - An additional 'special case' block was set aside for IXPs.

This all impacts, in a positive or negative way, how much future there is in our final /8 policy. I'd like to think that we made a well-considered decision back then, and if we're going to make a fundamental change like this one I'd like to see some numbers in an impact analysis. Based on current distribution, how much space do we anticipate will fall under the final /8 policy, how much of it will be allocated in /22 PA and how much will be allocated in /24 PI? Given the 'one size fits nobody' nature of the final /8 policy, this would be about the number of allocations/assignments done so far, not the size.

Personally I'm rather sick and tired of hearing people say 'yes, let's break IPv4 so we all MUST move to IPv6'. If you think this is good policy or even good engineering, please think again. It will make us end up with a broken internet that, surprise, we won't be managing any more. Breaking IPv4 might lead to increased IPv6 adoption, but not on the internet as we know it today. Everybody who wants to connect his organisation to the internet is going to need at least some IPv4 address space for the time being, so why screw it up for new entrants?

Finally, I would like to hear how this proposal correlates to the charge for PI space - the good old 2007-01 chestnut. For post-depletion LIRs, the grapes would be quite sour if one could pick up a quarter of the available resources for about one twentieth of the price. Should "Final /8 PIv4" have a separate price tag?

Best

Remco

This email is from Equinix Europe Limited or one of its associated/subsidiary companies. This email, and any files transmitted with it, contains information which is confidential, may be legally privileged and is solely for the use of the intended recipient. If you have received this email in error, please notify the sender and delete this email immediately. Equinix Europe Limited. Registered Office: Quadrant House, 4 Thomas More Square, London E1W 1YW. Registered in England and Wales, No. 6293383.

Remco,

Alright then, for the sake of argument I'll oppose until I see some convincing numbers. Back in the original last /8 discussion the rationale for choosing a /22 was that it would get us about 16k final allocations, or 1 for every NCC member and room for the membership to double in size.

we need to move away from this idea of how to expand the RIPE NCC membership and think more in terms of how to serve the RIPE community. There may be a good deal of overlap between these two goals, but they are not necessarily the same. There is no doubt that if address space can be assigned as PI, then this will reduce the amount available to LIRs and that this is not a good thing. On the other hand, who is to say that a LIR deserves IP address space more than an End User who needs a PI assignment? As it stands, the "last /8" policy makes a default unilateral judgement in favour of the LIR. This strikes me as being an extraordinary position to take.

Now, we have a number of new realities:

- final /8 applies to all v4 address space when it kicks in, including space that gets returned later; - It also applies to v4 address space that has not been allocated or assigned by RIPE NCC at that date; - An additional 'special case' block was set aside for IXPs.

The special case IXP block is neither here nor there. A /16 makes no difference in the grand scheme of things. What's relevant here is that as a result of 2011-03, the "last /8" policy would be more appropriately called "the dregs" policy - the ipv4 policies which apply to the last /8 will become permanent fixtures applying to all future address space after depletion. Perhaps they won't apply to a huge amount of address space, but there will be a constant and small turnover of address blocks reclaimed by the RIPE NCC for the foreseeable future. Turns out, this is a pretty fundamental change. We had, as a community, created a last /8 policy because we believed that there was something special about the addresses in the bottom of the barrel. Then we realised that there would be a permanent small trickle of addresses into this barrel and that there was really no such thing as the "last /8". In a roundabout sort of way, this policy floats the idea that the entire concept of the last /8 being special is slightly artificial, and that really they're not "special" addresses, they're just "addresses". Same as all the other addresses we've assigned, allocated and used all along. The only difference we've really made is that we've narrowed the mouth of the toothpaste tube so that more people might be able to get a taste. Taking a slightly different viewpoint, 2012-04 makes the last /8 policy more similar to the "run out fairly" model, except that instead of limiting on the basis of expected use within X months, we're putting some hard limits in.

This all impacts, in a positive or negative way, how much future there is in our final /8 policy. I'd like to think that we made a well-considered decision back then, and if we're going to make a fundamental change like this one I'd like to see some numbers in an impact analysis. Based on current distribution, how much space do we anticipate will fall under the final /8 policy, how much of it will be allocated in /22 PA and how much will be allocated in /24 PI? Given the 'one size fits nobody' nature of the final /8 policy, this would be about the number of allocations/assignments done so far, not the size.

yes, numbers would be interesting and I agree that the impact analysis should include some form of run-out projection of the type you're suggesting. I don't know if these numbers are going to make a fundamental difference to whether this is a good or a bad policy, but no harm will come from having them to hand. btw, do you have any criteria for evaluating these numbers if/when they are produced by the NCC?

Finally, I would like to hear how this proposal correlates to the charge for PI space - the good old 2007-01 chestnut. For post-depletion LIRs, the grapes would be quite sour if one could pick up a quarter of the available resources for about one twentieth of the price. Should "Final /8 PIv4" have a separate price tag?

This is quite a different discussion, and is part of the more fundamental issue of why we have PI and PA in the first place. If they're just addresses, why does the RIPE community want them packaged up and labelled with different colours? What do these colours mean, and how will this impact on ipv6 PI and PA assignment policy? How will it affect the RIPE NCC budget? How can we make the PA and PI charging schemes more aligned? Or even fully aligned? These are all interesting and relevant questions to both the RIPE NCC and the RIPE community, and I believe it to be critically important to the RIPE NCC that they be asked and answered soon. But they are out of scope for this policy proposal. Right now, we have a case of significant bias towards LIRs and this is what this policy proposal seeks to redress. Nick

* Remco Van Mook

This all impacts, in a positive or negative way, how much future there is in our final /8 policy. I'd like to think that we made a well-considered decision back then, and if we're going to make a fundamental change like this one I'd like to see some numbers in an impact analysis. Based on current distribution, how much space do we anticipate will fall under the final /8 policy, how much of it will be allocated in /22 PA and how much will be allocated in /24 PI? Given the 'one size fits nobody' nature of the final /8 policy, this would be about the number of allocations/assignments done so far, not the size.

Hi, So, by looking in today's delegated-ripencc-latest file... We need to look back on 105 days of delegations in order to as close as possible to a /8s worth of addresses handed out. The numbers are: Total number of delegated addresses: 17,156,520 Total number of delegations: 1,696 Total number of allocated addresses: 14,784,512 (86.17%) Total number of allocations: 701 (41.33%) Total number of assigned addresses: 2,372,008 (13.83%) Total number of assignments: 995 (58.67%) If I, eh, "simulate" the last /8 policy + 2012-04 by assuming that every assignment is a /24 and every allocation is a /22, we would need to look back 3071 days to get to the equivalent of a /8: Total number of delegated addresses: 16,779,776 Total number of delegations: 28,787 Total number of allocated addresses: 12,547,072 (74.77%) Total number of allocations: 12,253 (42.56%) Total number of assigned addresses: 4,232,704 (25.23%) Total number of assignments: 16,534 (57.44%) Note that neither of these filter away additional delegations to LIRs and EUs, something which won't happen under the last /8 policy. I don't make the claim that these numbers necessarily are meaningful or even relevant to the discussion, though. You be the judge of that...

Everybody who wants to connect his organisation to the internet is going to need at least some IPv4 address space for the time being, so why screw it up for new entrants?

Indeed. I suspect that this the exact problem 2012-04 aims to deal with, as the current last /8 policy *does* «screw it up for new entrants» [that aren't LIRs]. I have a couple of questions regarding the proposed policy, though (not directed at you Remco): 1) Under the proposed policy, would an LIR be eligible for both a /22 PA allocation and a /24 PI assignment, or is it either-or? (I think the proposal as it currently stands is the former, but is that intentional?) 2) Will ripe-530 section 6.10 (minimum assignment=/24 if multihoming) be in effect under 2012-04? Best regards, -- Tore Anderson Redpill Linpro AS - http://www.redpill-linpro.com