[ppml] Address Space versus Routing Slots

Tom Vest tvest at pch.net
Sun May 7 22:42:54 EDT 2006


On May 7, 2006, at 9:48 PM, Martin Hannigan wrote:

> At 5:36 PM -0400 05:07:2006, Jason Schiller (schiller at uu.net) wrote:
>> On Sat, 6 May 2006, Martin Hannigan wrote:
>>
>>>  At 08:17 PM 5/4/2006, David Conrad wrote:
>>>> Hi,
>>>>
>>>> On May 4, 2006, at 7:53 AM, william(at)elan.net wrote:
>>>>> I agree with Michael - IPv6 PI is not perfect but we desperately
>>>>> need such policy to help in adoption of IPv6 in the US - or  
>>>>> otherwise we will
>>>>> run out of ipv4 space before majority are ready.
>>>>
>>>> I'm curious: what do people think will happen when IANA  
>>>> allocates the
>>>> last unused /8?
>>>
>>>  Nothing. The commodity market for IPV4 will move above ground prior
>>>  to that and instead of faking asset acquisitions and lying on  
>>> applications,
>>>  we'll see blocks openly traded in commodity or auction markets.  
>>> Legacy
>>>  space will have value and people will exploit that. Real money will
>>>  exchange hands.
>>>
>>>  The registry/registrar mess is a whole lot of fun to watch.  
>>> Perhaps it
>>>  makes sense to work on the conversion now and let the market  
>>> decide when
>>>  it's time for v6 based on value of V4?
>>>
>>>  -M<
>>
>> Not that I disagree, but what will be the impact of the IPv4  
>> commodity
>> market in relationship to the routing table and the concepts of  
>> good IPv4
>> stewardship?
>
>
> Nothing, I suppose. Operators would not be asked to change policies.
>
> You'd add slots, but not at the rate of the feared v6 influx. Operator
> routing policy would effectively be an influence on the pricing so
> theoretically, operators could squeeze the prefix. This needs more
> analysis.
>
>> First off, it is likely that those with large amounts of legacy  
>> will sell
>> off small unused chunks of that space.  "So you need a /22 to turn  
>> up a
>> new customer, and I've got this legacy Class B.  I could sell you  
>> the .26,
>> .113, .117, and .118 subnets."
>>
>> This results in de-aggegragtion and many more specific in the  
>> routing table.
>>
>> This also makes the only justification for IP space how much money  
>> you
>> have.  This means poor companies (or countries) may be left out in  
>> the
>> cold increasing the digital divide, and spinning up the efforts of
>> WSIS and the UN.  It is also possible that rich companys might  
>> stockpile
>> the  remaining addresses.  What would happen if say Cogent bought  
>> up all of
>> the remaining space?  Can you see other ISPs hanging out the "no  
>> address
>> vacancy, please try Cogent" sign?
>
>
> I don't know that poor people would be up the creek. We don't know
> what the pricing would be. Could be. But...
>
> For example, if they want PI, they can go to the commodity market
> and buy it. If not, they can pay for PA space from their provider.

In some places with limited transport options, PA in nontrivial  
quantities (/24 or larger) cannot be had at any price.
In some places with limited transport options, PI (or even non-NIR  
allocated) will not get routed at any price.

Obvious implications -- a fair number of "poor people" are already up  
the creek, but for the most part only in places where other critical  
infrastructure bottlenecks already exist...

With new bottlenecks looming in more and more places, some attention  
to likely market outcomes is probably (strongly) warranted.

TV

> It may also go the other way where the price of a legacy /16
> isn't worth as much as people think.
>
> Regardless, as the intrinisic value rose, there would be
> ample reason to go to a V6 stategy since the amount of
> address space alone would theoretically keep prices low enough
> for a point of entry of almost anyone.
>
> BTW, our friend Aaron Hughes has an excellent "cheat sheet" on
> prefix lengths and recently included v6.
>
> http://www.bind.com/netmasks.html
>
> As far as the Internet have-nots, Africa has less than 2% of all
> distributed public facing root-servers. The continent with less is
> Antartica. (I made a preso related to this last week, as a matter of
> fact)
>
>
> -M<
>
>
>
>
>
> -- 
> Martin Hannigan                                (c) 617-388-2663
> Renesys Corporation                            (w) 617-395-8574
> Member of Technical Staff                      Network Operations
>                                                hannigan at renesys.com
> _______________________________________________
> PPML mailing list
> PPML at arin.net
> http://lists.arin.net/mailman/listinfo/ppml




More information about the ARIN-PPML mailing list