Broadening the IPv6 discussion

Thu Aug 29 20:39:19 UTC 2002

On Thu, 29 Aug 2002, Daniel Golding wrote:

> Hmm. I'm afraid that I have to disagree with just about everything you've
> said :) . I haven't seen any enterprise folks demanding v6 - If VOIP and
> PDA's (?) use up their IP addresses, they can easily ask for more. The more
> you use, the more you get. There is no shortage of v4 space.

As far as I know, we're still scheduled to run out of IPv4 address space
this decade. But it's anybody's guess if this is really going to happen
(even if you define "running out" as "too hard to manage" rather than
"nothing left"). Address usage will follow an S curve: slow start, then
steeper and steeper, until you come close to "everyone that wants one has
one" and then it levels off again. The question is: where on the S are we
now? There is something to be said for high (close to leveling off)
because pretty much anyone who wants/needs IP in North America and Europe
has it, but maybe we're still quite low, since lots of stuff that could
benefit from IP connectivity is still standalone. (And then there's the
rest of the world, of course.)

> Basically, major backbone networks will deploy v6 when it makes economic
> sense for them to do so. Right now, there is no demand and no revenue
> upside. I don't expect this to change in the near future.

The question is not if they're going to carry v6, because they already
are. The question is: will they do native v6, or tunnel it over v4? Since
next to none of the high end stuff can do native v6 at wire speed, it's
obviously still the latter now, but this is something that can change
relatively easy. There are already many signs of impending v6 adoption:
exchanges such as the AMS-IX are starting to do native v6, OSes have it
built in, router vendors are implementing it deeper inside the hardware
rather than at the main CPU level. However, noone is in a big hurry.
That's probably a good thing. When we really need it, IPv6 will be good
and ready.

> v6 is, currently, a solution in search of a problem. v4 space is being
> consumed slowly, but we are quite some time from a crisis. Of course, even
> when we "consume" all such ipv4 space, there are still expedients that can
> be used, including making v4 assets tradable and fungible.

The problem is not so much address space (you can run a fortune 500
company behind a single address with NAT) but routing. This is still a big
problem in IPv6 (as we're hoping to avoid the mess that is IPv4), but I
think we're getting closer to a solution.

Iljitsch van Beijnum