[arin-ppml] V6 address allocation policy
tedm at ipinc.net
Fri Jan 15 17:23:21 EST 2010
Davis, Terry L wrote:
> I've been a big supporter of IPv6 for a decade now since I was in the
> FTTH business for awhile in 2000-2001. Industry has spent an
> enormous amount in developing it both in network and in the end
> systems. And I still feel it has huge potentials to allow us to
> improve the Internet.
> But yet even with the globe rushing headlong toward the end of IPv4
> space, probably within 24 months, v6 is still barely crawling forward
> in deployments. It's not going into greenfields, startups, etc. It
> is still hard to find native v6 transport. I don't know of a v6
> network anywhere approaching even approaching 100,000 systems (I hope
> I'm wrong!) on the globe.
> Yea I finally realized in doing my Master's paper a couple years back
> that we had really screwed up by not defining a native way to allow
> v4 to v6 communications.
Not true. I used to run OS/2. Remember that? OS/2 Warp?
Well let me tell you something about transitions. IBM knocked
themselves out adding seamless windows support into OS/2 Warp.
They really wanted to be able to say that Warp ran Windows
better than Windows does. And they succeeded so well that
their software partners - like DeScribe - who for years ONLY
produced OS/2 versions of software, ended up going out of
business because all the Warp users out there simply used
their legacy Windows applications under OS/2 and never bothered
switching to OS/2 apps. Why would they, when Windows apps worked so
well under OS/2?
Some things call for backwards-compatibility. Some things instead
call for making it very painful for the customers so that they are
forced to spend money to upgrade - because their upgrades are for
the greater good of the community. The customers who refuse to
upgrade are then cast-aside, they are winnowed out. It may seem
unfair - but to this day there's still people out there who have
refused to give up their Commodore 64's and buy PCs. That is
just a fact of life with change. Some people refuse to accept it
and will just continue on with what they know - until they are
among a small minority, and then they die of old age.
Look at the HDTV business. We all know the US Government gave everyone
free converter boxes to get their crappy old TV sets to work
on HD. But, the US Government DID NOT pay for anyone to get a
brand new HDTV UHF antenna, even though millions of people were
running set-top rabbit ears, or VHF antennas on the top of their
roofs. And those millions of people were basically told you
go spend $35 on a new Channel Master UHF antenna and find some
handyman to climb around on the top of your roof and install it.
We aren't going to make the signal backwards compatible to your old VHF
antenna because we know damn well you wouldn't lift a finger to replace
We know that customers aren't going to spend money unless they
have to. Sometimes you just gotta be a hard-ass and don't give
them a choice to NOT spend the money. This is one of those times.
> As is, you basically have to open every v4
> app and re-write it to utilize v6;
> none of the existing transition
> technologies cover all the v4 to v6 communications scenarios. With
> this much installed v4, the cost of opening every existing app to
> change it to be dual-stacked is staggering.
That doesn't matter. All of those apps your talking about are
going to be obsolete in 20 years and replaced by new versions so
that staggering cost is going to be spent either way.
> We can argue endlessly about the risks of opening v6 address
> allocation policy but in the end, if we cannot get the Internet
> developers to utilize it, all the investment of the IT and comm
> vendors will be lost. One of the alternatives to IPv6 will win (geo
> routing, 5th octet, something-out-of-the-blue, etc) and all that
> investment in IPv6 and its potential enhancements to the Internet
> will be lost to us.
If you really think that an alternative to IPv6 has a chance then
where are all those startup software companies writing to one of
those alternative standards? Why isn't Microsoft pushing one of
out-of-the-blue laboratory curiosities implemented on Linux just
aren't going to make any difference. The future is IPv6 and
all the big players are betting on it, and the economic
situation in the world right now is not such that anyone is
going to put any real money into an alternative.
The question isn't whether it's going to be IPv6 vs some kludgy
IPv4 alternative. The question is going to be how far can we
stretch the IPv4 that we have.
It's been observed before on this list that most large networks
have very "loose" allocations. For example the standard customer
static IPv4 allocation is a /29 and a /30 on a point-to-point link to
that customer. In reality it could be a /30 and unnumbered on the
point-to-point link since it's almost a given that all of
the customers getting /29's are only using a single number.
And if you want to force the end user to use a /32 you can
run PPP right to their router.
I think most established ISP's are aware of this and figure they can
self-generate IPv4 for 3-5 years post-runout. Their feeling is
why should I kill myself trying to kick my peers asses
to get them running IPv6 natively, when I can do nothing and allow
all the deep-pocket startup ISP's out there who are flush with
VC funding and have no IPv4 stored up, to beat my peers for me. Then
once my peers are IPv6 native, I'll just switch it on and be gold.
It kind of sucks for the new guy on the block, but that is also
a normal characteristic of established markets.
More information about the ARIN-PPML