[arin-ppml] Draft Policy ARIN-2019-10: Inter-RIR M&A - Seeking Community Comments

hostmaster at uneedus.com hostmaster at uneedus.com
Mon Jul 15 17:26:55 EDT 2019


It depends on if this is PI or PA space.

In the case of PI space, you may be right, one route removed, one route 
added - net change likely zero.  However in the long run, PI space is 
going to make the IPv4 routing tables look small when everyone has their 
own route, rather than a combined route with your upstream in the DFZ.

In the case of PA space, the default route leads to the upstream of the 
block, and breaking out a portion to be routed differently clearly adds 
routes.

This is one place that the IETF has let us down. IPv6 was supposed to 
allow multiple routers, each with its own upstream. I do this, but it 
requires hacks that should not be needed.  According to the specs it is 
supposed to be possible to multihome in IPv6 using only your provider(s) 
provided PA space.  The problem is that many OS's do not deal well with 
more than one router, so people are getting PI space and using BGP 
sessions to get around the shortcomings of what was supposed to be 
automatic multihoming and renumbering.  When this is fixed, the need for 
PI space will be greatly reduced.

With the sheer volume of IPv6 addresses at some point the growth of the 
routing tables because of PI addresses will not be sustainable, and some 
providers might even black hole them.

I have had to renumber in the past, and at least in Linux, sed is your 
friend. It really is not that hard.

Albert Erdmann
Network Administrator
Paradise On Line Inc.

On Mon, 15 Jul 2019, Job Snijders wrote:

> On Mon, Jul 15, 2019 at 05:01:43PM -0400, hostmaster at uneedus.com wrote:
>> I understand that we allow this in IPv4 only because of the shortage.
>> Further, changing IPv6 addresseses is not as big of hardship as it was
>> in IPv4 land, since both networks can exist during a changeover
>> period. Also, each segment always uses a /64, allowing easy changes of
>> the first 64 bits with automated tools in most Operating Systems.
>> There is NO shortage of IPv6 addresses, so why should we cause
>> unneeded expansion of the routing tables just to prevent a single AS
>> from having to renumber their single IPv6 network?
>
> Can you demonstrate how the routing tables will expand? This "argument"
> has been brought up a few times, but it is not clear to me how an
> administrative transfer from one RIR to another RIR has anything to do
> with the BGP tables.
>
> Kind regards,
>
> Job
>


More information about the ARIN-PPML mailing list