[arin-ppml] IPv4 is depleted today - unrealistic statements about IPv6 inevitability
Iljitsch van Beijnum
iljitsch at muada.com
Wed Sep 3 10:10:20 EDT 2008
On 3 sep 2008, at 6:42, Stephen Sprunk wrote:
> it seems like
> every year I've been in the tech biz, someone is shouting that the sky
> is falling and we've finally hit the wall on Moore's Law, yet a year
> later some other person has come up with a magical solution that saves
> us all and puts that particular crisis off for another year or two.
The sky isn't falling but there's also no magical solution. We know
that if we increased the number of prefixes in BGP by, say, 2 or 4
times overnight, large parts of the internet would fail. We don't have
anything that keeps this from happening except the expectation that
tomorrow won't be _too_ different from today.
After the IAB routing and addressing workshop where Tony talked about
this stuff several people from router vendors came out and said they
can (some even do) build routers that can handle 2 million prefixes:
nearly 8 times the IPv4 table that we have today. Presumably, those
routers will be in place by the time we hit that many prefixes. But
all the lines keep going up, and the "bad" ones (that we don't
control) cross the "good" ones (that we do control) at some point.
This is not an immediate issue, but it does pose a problem in the
> I remember when folks were telling us x86 CPUs couldn't run faster
> 100MHz, and they had lots of great statistics too, yet now the same
> folks are talking about how the 4GHz barrier is insurmountable...
And during the last bubble people thought we'd never again have an
>>> But what sort of configuration work, software upgraded,
>>> complexities, flakiness, support calls etc. would be required
>>> for ordinary end-users to run their home networks dual stack
A local IPv6 router. Apple sells one that's also a pretty nice wifi
base station and a rather slow file server.
Plus, if you have Windows XP, a single command line command or
fiddling with the network settings a bit.
IPv6-only is harder.
> That ignores all the apps and CPE devices (mostly DSL/cable "modems",
> which have an integrated NAT/FW function) that don't understand IPv6.
> Even if upgrades were available, which in most cases are not, you
> have to get all the users to install them...
> Or, we could just get NAT-PT working. At least one vendor is already
> shipping it.
NAT-PT means that you run IPv6-only locally and you connect to IPv4
servers through a translator. This means you must be able to do DNS
lookups over IPv6 (which XP can't do and MacOS can't autoconfigure)
and have IPv6 connectivity, so your CPE must support it. In other
words: it doesn't solve any of the issues you mention in the previous
More information about the ARIN-PPML