[ppml] NANOG IPv4 Exhaustion BoF

[Scott Leibrand]

David Conrad wrote:

> As for unnecessary deaggregation, what is or is not necessary is likely 
> a matter of opinion.  I firmly believe ISPs will look after themselves 
> as they have done in the past and are doing so today by applying filters 
> if they feel their infrastructure is at risk.  In this particular 
> respect, we've been here before and some folks still have the T-shirts 
> (hopefully they've been washed). I'd be curious to understand why past 
> solutions would not apply.

The type of unnecessary deaggregation I'm trying to avoid is this:

Organization A's IPv4 needs are growing, but their budget is tight.  In 
order to meet this quarter's needs, they transfer a block from org B. 
Next quarter, they again need new addresses, so they transfer another 
block from org C.  Before the year is out, they have at least 3 extra 
routes in the routing table, which can't be aggregated, nor can they be 
filtered without breaking connectivity.  Rather, I'd like to require A 
to get a large enough block up front to meet their needs for 6-12+ 
months, thereby reducing the amount of deaggregation and the externality 
effects on the rest of the Internet.  (And if they really can't afford 6 
months of space, they can always lease PA space from their ISP or 
another LIR.)

>>> b) What tools exist (or can be expected to exist given reasonable 
>>> time/resources) to enforce that policy?
>>
>> The main tool is that, as the recognized authority in registration of 
>> IPv4 addresses in North America, recognition as valid of any transfers 
>> by ARIN has considerable value to both transferors and transferees.
> 
> So it would seem a core criteria in any policy would be to minimize 
> effects that would force folks to go elsewhere to have their transfers 
> recognized, no?

I don't think those effects have to be minimized, if the limitations 
help meet other goals.  But we do need to make sure that the costs of 
those effects is less than the benefits of using the white market.

>>> - the 6 month restriction could force folks to go outside the policy 
>>> in desperation (e.g., the amount of address space available via 
>>> transfers is likely to be hard to predict.  You could be in a 
>>> situation where at one point in time, the only option is a small 
>>> block even though you know it won't last 6 months.  What option do 
>>> you have?)
>>
>> You could get PA space from your ISP or another LIR.
> 
> I am assuming the folks most interested in getting address space will be 
> ISPs so they can continue adding customers.  Is the assumption of this 
> policy that the consumers of address space are end users?

Agreed.  Most end users will get their space from ISPs.  Most ISPs would 
probably prefer to "own" their own IPv4 space, but in a pinch they could 
get addresses from an upstream or a non-connected LIR (say a /8 holder 
who'd rather lease out space than transfer it).

> I fear the restrictions you are imposing will make it essentially 
> impossible to "do it right" and will result in folks with address space 
> finding other outlets in which to meet the needs of those who need 
> address space.  However, perhaps I misjudge the situation.

Can you identify which restrictions are onerous and would force people 
to go elsewhere?  We just submitted version 1.1 (which should be posted 
to PPML today), which allows deaggregation of transfered blocks to the 
level required to ensure adequate supply at smaller block sizes.

-Scott