Re: [arin-announce] IPv4 Address Space (fwd)

[Andy Dills <andy () xecu net>]

On Tue, 28 Oct 2003 Michael.Dillon () radianz com wrote:

> The bottom line is that there are three different models
> which may predict when we run out of IPv4 addresses. The
> models predict dates ranging from 2022 to 2045. None of
> the models predict an exact year, they all predict a range
> of 4 to 8 years and the above dates are the earliest and
> latest of those ranges.

Ok, so let's assume 2022, for the sake of argument. That is, after all,
nearly 20 years from now.

> > Does anybody honestly think companies will commit the capex needed to
> > implement IPv6?
>
> Yes, because IPv6 is merely and incremental improvement, not a grand
> elegant solution to world hunger like ATM. Look at how we managed the
> incremental step of adding MPLS to our IPv4 networks. It was fairly
> painless because it uses the same boxes, the same people and the same
> management systems. And over time, the pain of doing MPLS is reduced
> because the bugs get worked out.

Yes, but did MPLS require different ASICs?

> Similarly, IPv6 is an incremental change that uses the same boxes,
> people and management systems.

People need training (but not all that much), management systems need
rewritten (not majorly), and boxes need hardware replacements to forward
at line rate (CAPEX ALERT).

> In fact, if you've put MPLS into your core, you only need to worry about
> IPv6 at the edge from the PE router to the CE router because you can use
> 6PE. The capex is being spent anyway by upgrading boxes to meet capacity
> needs. You didn't notice it but the new core boxes are all capable of
> IPv6 with a simple software feature upgrade.

Yes, but there will always be this issue of billions of dollars of
exisiting, perfectly functional, unable-to-forward-v6-at-linerate routing
gear. If you have a router completely filled with attached customers, why
would you upgrade that router? You would buy another one for future new
customers, but not upgrade the existing one. The new one might forward
IPv6 at linerate, but the old one still doesn't, and there is still not
sufficient motivation to upgrade that old router.

> NANOG rarely takes the lead in new product development and driving
> market demand. Someone else will sort out that problem.

Yes, but the growing consensus among network operators is that IPv6 is a
waste of time and money, a technology that solved a problem that no longer
exists.

If we won't sign off on it, these "other people" won't even have a chance
to.

> I know that I said IPv6 is an incremental change, but the world that it
> enables is not incremental. Imagine 30 years from now where the majority
> of people in the developed world have full two-way voice, video, and
> data communications capability seamlessly integrated into their
> clothing, their vehicles, their workplace cubicles and their homes. X10
> is obsolete replaced by IPv6 over power networks and IPv6 over Bluetooth
> v.3. Networks are everywhere and it is common for even small devices to
> have multiple IPv6 addresses.  My belt (containing the cellphone
> transceiver) will have 20 IPv6 addresses in 20 different subnets
> corresponding to 20 VPNs. If you know about today's SIP networks, it's
> like having a phone number in INOC-DBA, FWD, SIPPhone, IAXtel etc.
> Except that these will be IPv6 addresses because they aren't for voice
> traffic. One of the 20 VPNs will be for a heart-rate monitoring service
> that coordinates with my gym and my personal trainer. Another one might
> be for an insulin level monitor that connects to my physician and
> pharmacy. The pharmacist will know when the insulin pack in my shirt
> collar will be depleted and will dispatch a refill to my home
> automatically.

Like I said, I don't think people will be all that excited about their
heart-monitor being reachable with a globally routed IP. People only want
to be connected to a certain degree.

Hell, there are people JUST NOW getting cell phones, and even more people
who will never get them. Most people aren't interested in being
"reachable" 24/7. Even more people aren't interested in having critical
functions rely on technical mumbo-jumbo when they have grown up taking
care of themselves just fine.

I think you're WAY overestimating our culture's thirst for technology.
As a society, we're still coming to grips with DVDs, MP3s, and cell
phones.

> That's the problem that IPv6 is intended to solve. Providers with
> foresight can begin the process of upgrading today so that when IPv6
> really is needed they will have a headstart. I'm not suggesting that
> anyone should rush IPv6 into production today, but everyone on this list
> should be running internal IPv6 trial networks to facilitate training of
> their personnel and to ensure that people have the experience needed to
> make rational and informed decisions about IPv6.

Getting ready for 2022 a little early, aren't we?

> In 1995, the Internet, a.k.a. IPv4 networks, took off with a boom and left
> the legacy telcos in the dust. If you want to recreate that, then keep
> your heads in the sand and other people will do IPv6 leaving you in the
> dust when critical mass finally arrives. It's that simple.

Yeah, those legacy telcos are so "in the dust".

That's why the only companies making real money these days are the
Verizons, Sprints, and BellSouths of the world.

Sorry Michael, but it's going to take more than fantasy about various
appliances having IP addresses (something IPv6 isn't even required for) to
convince the industry to spend billions of dollars to upgrade perfectly
acceptable routing gear.

Andy

---
Andy Dills
Xecunet, Inc.
www.xecu.net
301-682-9972
---