Re: [sig-policy] prop-091: Limiting of final /8 policy to specific /9
On 23/01/2011, at 8:35 AM, Owen DeLong wrote:
>
> There are three facts here:
>
> 1. IPv4 address space is finite and we are approaching the end
> of the free pools.
>
> 2. We do not know what needs will exist in the future and it may
> be prudent to set aside some amount of space in preparation
> for those needs.
We also don't know how exactly the industry is going to react to this. Clearly IPv6 as a product is a bit of a failure in terms of uptake (understatement of the week?). I've heard different voices suggest many different things, from the speculation of 'there will be a v4 trading system' through to 'we will add/remove/reconfigure technology to bypass the issue', even the more insidious 'we have the financial resources to buy companies that have the v4 we want'. So I personally don't know exactly how the industry will satisfy those v4 needs.
>
> 3. Space set aside for unknown future needs that is never used
> until after all needs are in the past is space wasted. Obviously
I'm not so sure that those 'unknown future needs' will pass by so quickly that we can't address them when they are identified in the policy development process. Certainly the PDP may not be as nimble as some would like, but it isn't a sloth either.
> to the extent we have that, we can say it was not prudent to
> keep that space in reserve.
>
> Unfortunately, balancing available resources between facts 2
> and 3 is the tricky part. Currently, the community has come to
> consensus that for APNIC that means setting aside an
> entire /8. There is good data to suggest that it might be more
The data provided doesn't, and can't, take into account how the industry is going to react. Who knows, maybe when the last 5 /8s get handed to the RIRs in the coming [days|weeks|month] the entire industry does a massive v6 deployment, or maybe completely changes their service model.
> prudent to reduce that set-aside to a /9. If you have good data
> to show that a /8 is more likely to be needed, let's see it.
I just don't see that we have the knowledge right now to make _any_ call. Surely that is reflected on the 'split' in RIR policies on the topic? For all the homogeneity I see in the RIR policies, this one stands out as being mostly dissimilar as globally there isn't a shared mindset. So not being convinced by any arguments for this right now I'd prefer to be conservative in terms of enacting the wrong policy due to ignorance. I'm quite sure many companies are worried about getting access to IPv4 cheaply, since the RIR system only costs the industry about US$70M/yr (based on RIR revenues) which really isn't that much compared to a IP wholesale fee (think $2/IP/yr) model. As much as all this is worrying (and indeed speculative) I suggest making a nice pot of camomile tea and watching how it unfolds first before making a change in policy.
Cheers
Terry