[arin-ppml] Policy Proposal 102: Reduce and Simplify IPv4 Initial Allocations

[Ted Mittelstaedt]

William Herrin wrote:
> On Fri, Nov 6, 2009 at 3:58 PM, Ted Mittelstaedt <tedm at ipinc.net> wrote:
>> William Herrin wrote:
>>>> 1) Makes moot whether the requesting ISP is multihomed or not, with
>>>> this policy change all initial ISPs request under the same minimums.
>>> I disagree with offering small PI blocks to entities which are not
>>> multihomed.  While any registrants benefit from receiving a PI block,
>>> only multihomed entities do so without creating new overhead cost for
>>> everybody else.
>> Multihomed entities are already obtaining larger allocations
>> and subnetting them, as I stated in the Rationale.
>>
>> Your making a claim here that the simple property of
>> being multihomed somehow acts as a deterrent to an org for
>> subnetting, and splitting their advertisements, thus
>> creating new overhead cost for everyone else.
>>
>> The fact is that entities that are NOT multihomed have
>> far more incentive to advertise all their space as a
>> single block, with the lowest overhead to everyone else -
>> and entities that ARE multihomed have an incentive to
>> split their advertisements - and create more overhead cost
>> for everyone else.
>>
>> In short, logic seems to indicate that it's completely opposite from what
>> your suggesting.
> 
> Hi Ted,
> 
> That's why I'm an engineer and not a philosopher. Logic alone leads to
> all sorts of falsehoods.
> 
> The above, for example, is a straw man argument. A single homed entity
> can't engineer traffic to take one path or another since there is only
> one path. Hence they won't tend to disaggregate for TE like a
> multihomed entity might. Hence multihomed entities are more likely to
> disaggregate whatever sized block.
> 
> But that's entirely beside the point. Single-homed entities require
> zero routes in the DFZ to function as designed on a day to day basis
> while multihomed entities require at least one. That was the point.

Except that there are significant business/economic problems with it
working the way your describing.

While it is true that a single homed entity that qualifies under 4.2.2.1 
for a /20 CAN have their single upstream feed advertise that /20 under
the upstream's AS, in practice this is almost never done.

One big reason ISPs obtain portable numbering so they have the freedom 
to be able to tell their upstream to go stuff themselves if they decide 
the upstream is screwing them over, without the fear of renumbering 
pain.  Thus if they are going to go through the pain of a renumber into 
a portable /20 the very last thing they are going to do is just turn
that /20 over to their upstream to aggregate in it's own advertisements.

Instead they are going to get their own AS and do BGP to their upstream.
Thus, consuming "at least one" route in the DFZ.  This allows them
to shift upstreams within minutes.

> Not talking convenience here, not even typical practice, just basic
> minimum function. Excluding single-homers discourages the group whose
> systems will still work right from announcing any routes at all while
> multihomers will announce as many if not more routes regardless, just
> not using addresses directly allocated from ARIN.
> 
> To be completely honest, it's no skin off my back either way. When the
> DFZ passes 1M entries, I only have to replace a few $1500 routers. I
> suspect, however, that the folks who spend half a million per router
> may see things from a different perspective... one that would prefer
> to see few folks in the "zero routes required" category actually
> showing up in the table.
> 

Bill, this proposal only lowers the bar a little bit, that bar is still
there.  They still have to show efficient utilization of at least 400
IP addresses.  And, if their goal is to renumber into a portable ONLY so 
as to "never have to ever do another renumber ever again" well that 
isn't going to exist, either.

A small ISP that has friendly relations with their upstream really
doesn't have much incentive to request numbering under this policy for a 
  "provisional" allocation.  They have much more incentive to simply 
keep plugging away, growing their business, until they qualify for a /20,
since that's the minimum allocation that's a permanent allocation,
rather than a provisional allocation.

It's only the small ISP's who have upstreams who are screwing them
over, either by NOT supplying products (like native IPv6) that
they want, or by raising their rates far in excess of the market
rate for their area, who are going to be wanting to renumber into
a portable allocation that they can move elsewhere.

Look, you know that a small ISP that cannot afford to multihome
isn't going to be able to afford move to a different upstream, if
that upstream has supplied them numbers.  A renumbering project
can take months, and they would have to maintain 2 feeds during
the transition, they can't afford that.

If this policy is put into effect I would guess that it's very
existence will act as a check on the large upstream feeds from
jerking their small ISP customers around, and that will very likely end
up helping to discourage the smaller ISPs from trying to obtain
portable numbering in the first place.

> If I hear different from the likes of Verizon and AT&T, I'll drop my objection.
>

Would you be willing to drop your opposition if you DON'T hear 
opposition from the likes of Verizon and AT&T?  After all, if they 
figure it isn't going to make much difference to the DFZ, then they have 
every incentive to remain neutral.

Ted

> Regards,
> Bill Herrin
> 
>