[arin-ppml] Some observations on the differences in the various transfer policy proposals

[Tom Vest]

On Oct 20, 2008, at 12:25 PM, David Williamson wrote:

> On Sun, Oct 19, 2008 at 06:47:46PM -0700, Scott Leibrand wrote:
>> Or, to put a more practical face on the same question: how do you  
>> propose
>> that the industry deal with the deaggregation that will result from  
>> the
>> widespread transfer of small netblocks (as allowed under prop-50)?
>>
>> I agree that restricting deaggregation through regulating access to  
>> the
>> registry will not be 100% effective, but it seems more likely to be
>> effective than any alternative I've seen so far.
>
> I know many or even most of us wear multiple hats, and routing table
> bloat is a very serious concern for us.  Depsite that, I'm still
> wondering why we spend so much time trying to prevent it via address
> policy.

Hi David,

That's the great thing about an operating environment that "works" --  
it's largely invisible/transparent to its beneficiaries.
But here's why we spend so much time doing it.

For the sake of argument, imagine that in most places the absolute  
success rate of first-time RIR "initial allocation" requests is around  
33% under most circumstances. Presumably, most aspiring new routing  
system participants that don't get approved initially take another  
turn at bat, perhaps with a better clarification of need, or with a  
lower need calculation. A few may give up, but presumably most get it  
right and eventually appear as their own PA or somebody else's (PI)  
routing tale entries. However, those that miss the mark first time  
take longer to appear there, and often appear with a somewhat smaller  
footprint.

Now imagine that hostmasters within ISPs have approximately the same  
"instant gratification rate" -- if not then you get more frustrated ex- 
customers arriving that the RIRs' doorsteps, or accelerated growth  
rate for subsequent allocation demand, which may also prompt RIR  
hostmasters to inquire about the effectiveness of ISP-level  
conservation practices.

The first line of evaluation determines how many PA and PI recipients  
there are, and how many more join there ranks every year. The second  
line of evaluation determines how often each is returning for a second  
and subsequent PA allocation (non-aggregatable in most regions), and  
also affects the frequency with which new, non-aggregatable PI seekers  
appear.

There's no real way to know the absolute rate of demand for new entry,  
as defined above, because we're just now reaching/exceeding the peak  
demand levels that existed at the time of the bubble. The Internet's  
much bigger now, however, so it looks much less bubble-like this time  
-- but whether or not this is a cyclical phenomenon, or the  
continuation of a single (albeit interrupted) growth trend is  
anybody's guess. But unless you imagine that the world's supply of  
aspiring ISPs and network-insourcing enterprises is almost permanently  
exhausted, maybe it doesn't matter.

So, what would have happened under a different allocation arrangement  
that made it twice as easy to secure an initial allocation? Maybe  
there would now be twice as many routing system participants, most of  
which would be facing twice the competitive pressure of the kind that  
results in PA prefix-level deaggregation. Alternately, what would have  
happened if half of the est. 3,500 or so new routing system  
participants that joined our ranks last year had been absolutely/ 
permanently turned away? How many national regulators would have  
opened up antitrust investigations under those circumstances?  
Alternately, how many would be preparing to partition their own  
economies and instituting their own, "fair" address delegation regimes?

The system that we all collectively maintain has been successful to- 
date because it has successfully balanced these requirements.

> We've been only marginally effective at it, as Geoff notes.


Geoff and others assume that PA prefix-level deaggregation could have  
been or should have been stopped somehow by the RIRs, even though the  
RIRs are basically just administrators of community-defined policies.  
They insist that this inflation management should have been successful  
despite the universal assertions of commercial independence -- and  
also assume that it would neither have resulted in a huge increase in  
the demand for PI prefixes -- leading to the same result -- or to  
calls for external intervention. Such complaints should sound familiar  
-- they're identical to the claims that some bank executives have made  
recently, that they're merely innocent victims of bank deregulation,  
and the commercial freedoms which they themselves demanded.

TV


> It seems to me that the ability of a given allocation or assignment to
> get into a global routing table (whatever that may mean at the time)  
> is
> entirely not our (ARIN's or other RIR's) problem.  As the ARIN NRPM
> notes:
>
>> Provider independent (portable) addresses issued directly from ARIN  
>> or
>> other Regional Registries are not guaranteed to be globally routable.
>> Therefore, ISPs should consider the following order of priority when
>> requesting IP address space:
>>
>>   * Request IP address space from upstream provider
>>   * Request IP address space from provider's provider
>>   * Request IP address space from ARIN (not guaranteed to be  
>> globally routable)
>
> The language is a bit outdated, but the intent is clear.
>
>> From my point of view, good stewardship of the address space means
> conservation and efficiency.  That's almost diametrically opposed to
> preserving the routing table...but again, that's not ARIN's problem.
> It's a problem for the operator community to sort out, via new
> technology (RRG folks, save us, please!), or monetization of routing
> slots, or value judgements about specific routes (if a root server was
> anycast in a /28, would you route it?).
>
> Again, many of us live in both worlds, so it's difficult to separate
> those conflicting needs, but we need to at least acknowledge the
> problem.  More on topic, I do think we need to consider Geoff's point
> about what regulatory role we see for the RIR system in the future.
> I'm not convinced that abstaining from any regulation is a good idea,
> since we do have a bit of a lever to yank on, but heavy-handed
> regulation will almost certainly result in alternative registries
> getting setup.  While alternative roots for DNS are mostly an amusing
> and harmless joke (to me, at least), alternative authorities for who
> uses what IPv4 space seems like a path to chaos.
>
> For my part, I think the level of regulation imposed by the current
> *allocation* policies is sufficient.  Adding extra verbiage to a
> transfer policy may only cause confusion.
>
> Perhaps we could see a copy of the open mic slide from rs posted here?
> That seemed like a more rational way to move forward with a transfer
> policy.
>
> -David
>
> _______________________________________________
> PPML
> You are receiving this message because you are subscribed to
> the ARIN Public Policy Mailing List (ARIN-PPML at arin.net).
> Unsubscribe or manage your mailing list subscription at:
> http://lists.arin.net/mailman/listinfo/arin-ppml
> Please contact info at arin.net if you experience any issues.