[arin-ppml] Statistics request regarding new entrants (was: Re: Stats request)

[Owen DeLong]

On Nov 27, 2013, at 16:56 , mcr at sandelman.ca wrote:

> 
> Owen DeLong <owen at delong.com> wrote:
>>> For ISPs, that is very likely the case (due to policy.)  Note this
>>> highlights the potential issue for new ISP requesters, in that it is
>>> going to become (or is becoming) increasingly difficult to qualify if
>>> upstreams do not provide sufficient PA IPv4 space to smaller ISPs such
>>> that they can qualify for a direct ARIN allocation later.
> 
>> The word you are looking for there, John, is "has become". We've
>> already seen multiple reports from members of the community that they
>> are deadlocked on this issue because their upstream will not give them
>> more space and they don't have enough space from their upstream to
>> qualify through ARIN.
> 
> Yes, but I do wonder whether having these ISPs *plan* to deploy XLAT464 to
> existing customers might provide:

First, we have struggled to keep policy agnostic and I don't believe that making
policy to force new entrants to inflict translation services on users is a particularly
good policy choice.

Second, even if we choose to make such a move (which I think would be grossly
unfair among its other flaws), it would have to be translation technology agnostic
at least.

>  a) enough breathing room to demonstrate need.

I don't see how.

>  b) we could then count the IPv6 networks.

This would require much more complicated changes to policy than anything that
is currently being proposed and I'm not sure what the perceived advantage would
be.

> 
>>> That's not readily apparent, but the number of IPv6 allocations to
>>> ISPs in total so far this year is 237
>>> <https://www.arin.net/knowledge/statistics/index.html> so it cannot be
>>> more than about 60% of new requesters along requesting IPv6.
> 
>> If they are new entrants, wouldn't it be fairly easy to look at whether
>> or not their ORG-ID has received IPv6? Since "have IPv4 from ARIN" is
>> pretty much automatic qualification for some amount of IPv6, I would
>> say anyone who requested IPv6 and is an IPv4 new entrant likely
>> received it to the point that any error from that assumption could be
>> considered statistical outliers.
> 
> received it does not mean deployed/advertised it and used it.

Sure, but ARIN can't really control that and has limited ability to influence it.

The closest ARIN can come is measuring what ARIN has done in terms of
allocations and assignments. I will point out that allocating or assigning IPv4
also doesn't necessarily mean that it got deployed or advertised, either.

> 
>>> ARIN has no visibility into any of the above, and note that per NRPM
>>> 4.2.2, standard ISP allocations are not required to return their
>>> blocks from their upstream ISP (those receiving under multihomed
>>> policy are required to return)
> 
>> The return requirement is primarily enforced by the fact that ARIN will
>> not issue additional resources unless/until the block(s) is/are
>> returned.
> 
>> What the upstream ISP does with the returned blocks, OTOH, would be
>> almost impossible to scrutinize and I suspect there are likely as many
>> different answers as there are ISPs receiving returned space.
> 
> Why can't ARIN *ASK* in the situations where the block is returned?
> We ask to see it returned before the subsequent allocation.  Upstream doesn't
> need to answer.

I'm not sure I understand your question here. Who are you seeking for ARIN to
ask what question?

We ask the downstream to return before we will grant the downstream a subsequent
direct allocation. The upstream isn't really involved other than (at least theoretically)
receiving the return.

In my experience there is a wide variance in the diligence of upstreams in processing
these returns which ranges from an extreme of not even noticing that the space was
returned and only removing it from SWIP or RWHOIS under substantial pressure from
the (often now desperate) downstream that long since returned the space to the other
extreme of placing it readily into their free pool and allocating almost immediately to
new customers with all manners of variation between those two extremes.

In my estimation (which is fairly close to a SWAG but based on significant industry
experience and conversations with a number of providers over many years), there
are far more providers somewhere closer to the extremely inattentive end of the
spectrum than the aggressively recycling end of the spectrum.

Owen