[ppml] Proposed Policy: IPv4 Countdown

[Ted Mittelstaedt]

>-----Original Message-----
>From: Howard, W. Lee [mailto:Lee.Howard at stanleyassociates.com]
>Sent: Saturday, March 17, 2007 9:26 AM
>To: Ted Mittelstaedt; ppml at arin.net
>Subject: RE: [ppml] Proposed Policy: IPv4 Countdown
>
>
>Wow, this post got very long.  I apologize to everyone; I'm
>trying to trim as I go, without losing context.
>
>I'm not sure how many more ways I can say the following:
>1.  Send a proposal.
>2.  IP numbering policies are determined on this list, and
>informed by public policy meetings.
>3.  ARIN is us.
>4.  Please send a policy proposal.
>
>I will continue correcting factually faulty assertions about
>the ARIN organization (community, Advisory Council, Board, 
>staff), process, and fees.  I will try to refrain from further
>posting about style, conspiracies, or the intent of industries
>or other organizations.  The AC should continue extruding
>clear policy proposals from the email threads.
>
>
>> -----Original Message-----
>> From: Ted Mittelstaedt [mailto:tedm at ipinc.net] 
>> Sent: Wednesday, March 14, 2007 9:48 PM
>> To: Howard, W. Lee; ppml at arin.net
>> Subject: RE: [ppml] Proposed Policy: IPv4 Countdown
>
>> >Ted, I can't tell from your tone whether you're trying to 
>> rant or make 
>> >constructive contributions.
>> >
>> 
>> Well, I honestly consider the issue of IPv4 runout to be a 
>> lost cause, since as I already mentioned, too many deep 
>> pockets have a vested interest in making the transition as 
>> uncomfortable as possible.  In short, "they" want to switch 
>> the Internet over to
>> IPv6 and are simply not interested in any proposals that will 
>> extend IPv4.
>
>"Rant" then?
>
>> I personally believe that with effort on the number 
>> registries part, we could extend the use of IPv4 on the 
>> backbone until long after everyone on the list here has 
>> retired.  But I see no real interest in doing this among any 
>> of the major players.  Instead all "they" care about is 
>> making everyone buy new hardware and renumbering.
>
>Who are these "major players?"
>Telcos?  Cable companies?  Router manufacturers?  All of the
>above?  
>Since you say "buy new hardware" I'll assume you mean router
>manufacturers, but we've heard loud concern from a major ISP
>that the router people aren't provided boxes to scale for 
>five years.  
>

There was just recently a long annoyed thread on the Cisco
mailing list about this topic.  Many older 7206VXR's have
NPE300s in them and Cisco just stopped supporting that CPU card,
right BEFORE correcting a long running bug in Service feature
set that prevents people from running IPv6.  So ISP's who don't
upgrade and who run IPv6 on this platform are left in the
proverbial cold if a problem happens.  People were angry since
Cisco had promised IPv6 support for many years and the bug that
was fixed had been open for many years previous to end of support.

That is just a simple example.  More apt is the many ISP's who
gave routers for free or "rented" routers like Cisco 2500's, 1600's
and 1700's to T1 customers, a few years ago that was a fad
I think, (at least around here it was) and now those devices
need hardware ram and/or flash as well as service updates to
run IPv6 

As far as scaling that is a different issue and I think depends
much on how the switchover from IPv4 to IPv6 is managed on the
Internet.  Right now a full BGP table with 2 views consumes around
170-200K of ram in a Cisco, there are a lot of Ciscos out there
with NPE cards that are maxed at 256MB of ram, so if your talking
doubling the BGP route table with IPv6 or with many /24 advertisements
then people will have to upgrade.

These are all end-node AS sites, as I would presume that a large
network would be running a Juniper or a much larger Cisco that would
have different scaling issues.

>> But, I do think it is facinating to 
>> watch someone try to change the status quo.  Sometimes, they 
>> even succeed in doing it.
>
>This is frequently the case at ARIN.
>

>
>Did you see the transfer request?
>
>Did you see any subsequent applications for additional address space?
>

We were an old MCI customer and yes, all our old numbers went from
MCI to SAVVIS about 9 months after MCI disconnected from us. 

We did ask ARIN to be assigned those, we were told the entire MCI thing
was a mess and nobody knew what was going to happen.  So we got
a fresh allocation.  In the long run renumbering helped us and ARIN
was definitely kind to us and didn't jack us around, so I have no
complaints there.  But from a theoretical standpoint SAVVIS could
have got the block we were allocated instead of the MCI numbers,
and then we wouldn't have had to renumber.

That is why I used that example, as I had first hand knowledge of
the events.  It might have been a special case for ARIN.

>> But, in reality, since the subsidiary is a separate 
>> corporation, even if only on paper, the parent can merely 
>> continue to use the subsidiary's old name when dealing with 
>> ARIN and thus avoid any existing rules such as 8.1 Transfers 
>> that you cited above.
>
>A separate corporation "only on paper" is a separate corporation.
>I don't know how else to define it.
>

Exactly my point.  I don't know of any way to make policy around
this problem, either, other than to give a reclamation staff more
authority to shift things around.

>
>Come to a meeting, meet the IP Analysts for the large ISPs, and ask
>them why they don't renumber.  Look at the list of attendees, and
>ask them (you'll have to find their contact information another
>way).  Ask the Whois Technical POCs.
>

I know why they don't renumber, it costs money and is disruptive.
You have to take it as a given that very few -want- to renumber.
There are many more reasons for not renumbering than for renumbering.

> 
>> Also there is another reason that not many people have 
>> switched to IPv6 and that is, since it is costly and disrupts 
>> customers, if an ISP devotes resources to doing it, the ISP's 
>> customers that are affected are going to wonder why they 
>> simply don't just go to a competitor ISP that isn't requiring 
>> them to do all this IPv6 stuff.  In short it is a competitive 
>> advantage to NOT change your network and disrupt your 
>> customers, yet have everyone else change their network and 
>> disrupt their customers.  This creates a condition where all 
>> of the ISPs facing this would much rather have everything 
>> disrupted at one time.
>> 
>> If for example IANA announced that on January 1st 2012 that 
>> there would be no more IPv4 traffic allowed on the global BGP 
>> network, 
>
>IANA does not control the "global BGP network".

I disagree.  Let's assume that over the next 15 years more and
more networks switch over to IPv6.  let's assume that America
Online announces that in 2015 they will no longer accept e-mail
from IPv4 networks.  I know that would cause many ISPs to convert
even if they were fighting it because their customers would
insist on them doing it.

By 2019, there are so few IPv4-only networks that the costs of
keeping track of IPv4 exceed the minor amount of money the RIR's
are still getting from IPv4-only registrations.  Let's assume
that the RIR's announce that due to this they will no longer track
IPv4 allocations as of 2020.  People could still advertise them if
they wanted, but they wouldn't be tracked, so nothing would prevent
someone from just using any old IPV4 numbers they wanted.

I think this would be the effective end of IPv4 in the so-called
"global BGP network"

In other words, if IANA were to decide 2010 was "flag day" they could
merely announce that all IPv4 allocations as of that date were 
unmaintained, non-tracked, and shut down whois queries for them.
At that time, no production network that
was fiscally responsible to it's customers would date to remain IPv4
only.

>
>I'm not sure.
>You assert:
>- Router vendors will only support IPv6 in OS versions that require
>more flash/memory than recent chassis can hold, to force upgrades.
>- ISPs won't implement IPv6 until forced to do so, so they can 
>delay the expenses as long as possible.
>
>You conclude:
>- None of the above are interested in proposals which would delay
>the exhaustion of IPv4.
>

That's the gist of it, yes.

>I will counter-assert that router vendors and large ISPs do not
>control the public policy process.  They can send email and speak
>at the meetings only as much as anyone else.  The only argument I
>recall seeing against some proposal you continue not making is 
>that reclamation wouldn't be worth the effort.
>
>
>Really, send a proposal.  It may not be adopted, but it definitely
>won't be if you don't send it.
>

Very good, then.  I will predict it will be shot down, but we
will see.

Ted