[address-policy-wg] About the /22 allocation limitation

As a general point, I don't support micromanaging and squeezing IPv4 to the nth degree, be it company or RIR level, only because it inhibits the overall development of IPv6. 

We know the sustainable growth of the internet, and internet companies, relies on the adoption of IPv6. Of course certain budget/resource restrictions may slow some companies/networks from launching v6, for which time they may indeed require some additional v4 space. But constantly managing v4 in an ambiguous state is more costly, not just in the long run. 
Harvesting v4 should be seen as a stopgap, a temporary workaround. We, as a community, should concentrate our efforts on enhancing and encouraging the deployment of IPv6. 

Rgds, 
James

-----Original Message-----
From: address-policy-wg-bounces at ripe.net [mailto:address-policy-wg-bounces at ripe.net] On Behalf Of Sander Steffann
Sent: 11 April 2014 13:44
To: Dpto. Datos Television Costa Blanca
Cc: Tore Anderson; address-policy-wg at ripe.net Working Group
Subject: Re: [address-policy-wg] About the /22 allocation limitation

Hi Daniel,

Playing devil's advocate a bit. 

> Lets think the reserved pool grows enought. Lets say reserved pool gets arround a /10. There should be a policy that says:
> [...] When the reserved pool reachs the number of "N" millions of IP [...] a LIR that prove they need the space, will recieve a /X space

[...]

> In general terms, what I mean is. Lets give the chance not only to new LIRs but everyone to get more IP space, IF THEY REALLY PROVE the need.
> I didnt say anything about minimum or maximum allocation, and if I said something that mean that, sorry.
> I also said to change minimum allocation from /22 to something little, for not exhausting the reserved pool in "2 days" but seems to not be the right decision due to the routing table grow.

I am looking at some research on RIPE labs: https://labs.ripe.net/Members/kistel/content-ipv4-address-allocation-rates, and especially this graph: https://labs.ripe.net/Members/kistel/images/userfiles-image-rir-speeds-201003-permonth-absolute-jagged(1).png. It only goes up to 2010, but the trend of allocating 0.2 of a /8 per month seems to be stable over many years. And in those days there already was a strong need-based policy, so all that address space was really needed.

You suggest to let the reserved pool grow to a /10. That would probably take quite a long time (it has been growing now for 1,5 years and we haven't reached a /10 yet). A /10 is only 0.25 of a /8. If the trend would continue it would mean that in between 1 and 2 months the /10 would be used up. So, if we allocate addresses like we used to then we would wait (let's say) 2 years, followed by 1 or 2 months of allocating addresses, then waiting 2 years again etc.

And this assumes the allocation rate of the period when there was no rush. In the current situation people would know that they only have a tiny window of opportunity. I therefore suspect that the /10 would be gone in less than a week. Some people get lucky and have a larger number of IPv4 addresses, while the rest of the market would not get anything. This feels a bit like a lottery, and the economic imbalance created by it might be significant.

So to make something like this work you would need a different policy. Some things you can think about:
- stronger need required (but in a shortage situation there is lots of need)
- maximum allocation sizes (which means people get a bit more, but still less than they need)

This is difficult to get right. Try playing several scenarios in your mind, trying to look at the allocation speed, the potential loopholes to game the system, the resulting economic balance etc.

> Of course, Im always talking about the reserved pool, not about the last 185/8 pool.

Ack

> Im really new to LIR and to this mailing list. For that im writting here to read all of you, your experiences and your opinion, and then if possible, do something.

Very much appreciated! This is however a topic that has been thoroughly discussed in the past, and we know how difficult it is to make good policy considering the IPv4 runout. The current final /8 policy is the best this working group has come up with. While lots of people would be very happy if they were able to get more IPv4 addresses it will be difficult to make that happen in a fair and useful way.

But it is always good to look at alternatives every once in a while. Thank you for that!
Sander