[arin-ppml] Internet means IPv6

Lee Howard spiffnolee at yahoo.com
Thu Dec 30 16:14:10 EST 2010

> >> Why is putting 100 legacy machines on the  public network of less value 
> >> putting 100 new machines on the  public network?
> > What public network?
> The public IPv4 network that  will continue to exist for years, perhaps 
>decades. But certainly for another  5-10 years.

It's not a public network.  It's a large set of private networks that agree to 
interconnect using standard protocols.  The conditions for use of one of those 
protocols is changing, and interoperability will be limited to the extent 
network operators disagree on protocol implementation.  In other words, run your 
network your way, but don't assume you can connect to the other guy.

> > If you want to provide a bridge for a PC with arcnet and netx.com, so it can
> > ping  your token-ring OS/2 machine, or get finger'd by your ATM Irix box,
> > use  rfc1918.  If you want "Internet access," (with a capital I) that term  
> > soon mean IPv6.
> Actually, it will soon mean "IPv4 and IPv6"...  too bad it hasn't meant that 
>for the last couple of years.
> If I come to  an ISP in 2013 with an AS number and my own IPv4 addresses, I 
>expect they'll be  able to route them and provide significant connectivity to a 
>large and still  extant network.

"Significant" is a huge step down from where we are now.  It may be acceptable 
for your needs.

> >   In 2012, there will be no expectation of  connectivity over
> > IPv4; if devices are unavailable over IPv6, it will be  the responsibility of 
> > laggard to upgrade.
> Only if they  need those to be reachable by IPv6 endpoints. In 2012 there will 
>still be lots 
> of IPv4 endpoints talking to lots of other IPv4 endpoints, and neither end will  
>need to upgrade.

Sure--if you control both endpoints and a significant portion of the connections 
between them
(enough to preclude CGN), then you can use IPv4 or whatever protocol you want.  
But you
should not assume you will have the same breadth of connectivity you do now.  

that's fine--you know what your hosts do.

> If an organization gets another /24 worth of IPv4 space  via transfer in 2012, 
>they'll be 
> able to announce it and hook up another 250  endpoints and talk to and from all 
> other IPv4 endpoints. And this will  probably be true in 2013, and 2014, and 
> even in 2024.

Sure, if you only need to talk to/from IPv4 endpoints.
And if you don't rely on IP geolocation for advertising services.  And you log 
address, port, and NTP timestamp on your server logs, so you can provide those 
law enforcement when you're attacked.

> > Or does it make sense for ARIN's analysts to be able to apply a  set of
> > objective criteria for evaluating requests?  That has been  ARIN's practice.
> That *does* make sense. What doesn't make sense is having  "applicant has 
> out how to deploy enough IPv6 to impress the analyst" as  a criteria for 
>whether IPv4 
> space is needed.

Are you arguing about the definition of "enough"?  
Would it make more sense for the application to say, "I've considered all of the 
cases, and I don't need connectivity to and from IPv6 hosts, just IPv4."


> Matthew Kaufman


More information about the ARIN-PPML mailing list