IPv6 mistakes, was: Re: Looking for an IPv6 naysayer...

Owen DeLong owen at delong.com
Fri Feb 18 14:48:12 UTC 2011


On Feb 18, 2011, at 2:50 AM, Iljitsch van Beijnum wrote:

> On 17 feb 2011, at 17:35, George Bonser wrote:
> 
>> Considering v4 is likely to be around for another decade or two, getting
>> Class E into general use seems easy enough to do.
> 
> You really think people will be communicating over the public internet using IPv4 in 2031?
> 
For some minimal definition of two endpoints both of which are IPv4, sure.
It'll be across 4in6 tunnels or something like that, but, I'm sure there will still be die-hard
legacy systems doing that in 2031.

As to whether IPv4 will still be generally routed on the internet? I actually suspect that
will end before 2021 and might start winding down as early as 2014. Many people
think that is overly optimistic, but, I look at the scaling problems IPv4 routing will face
in a post depletion world and I suspect the motivations to deprecate IPv4 will come on
strong and fast as a result.

Before you ask, no, I'm not going to promise to eat my column. (Hi Bob!)

> It will take a long time before the first people are going to turn off IPv4, but once that starts there will be no stopping it and IPv4 will be gone very, very quickly.
> 
Define long time. I'm thinking 3 to 5 years, maybe.

> (Of course there will be legacy stuff, just like some people are still running IPX or AppleTalk today. I'm talking about the public internet here.)
> 
> Today people are complaining how annoying it is to have to learn new things to be able to run IPv6, but that doesn't compare to how annoying it is to have to learn OLD things to keep running a protocol that is way past its sell by date. I still need to teach class A/B/C despite the fact that CIDR is old enough to drink in most countries because without knowing that you can't configure a Cisco router. That's annoying now. Think about how insane that will be in the 2020s when the notion of requesting IPv4 addresses from an RIR is ancient history and young people don't know any better than having a /64 on every LAN that is big enough to connect all ethernet NICs ever made.
> 
I am not convinced you can't configure a cisco router without knowing about classful addressing. True, you
will have to understand classful routing for the way Cisco displays routes to make sense to you, but, if you don't,
all that happens is you wonder why they display things so strangely, grouping these octet-bounded collections of
routes.

Owen






More information about the NANOG mailing list