New ARIN ipv6 allocation policies

Doug Barton dougb at
Sat Sep 2 01:23:58 CEST 2006

David Conrad wrote:
>> But if the choices are, "Only adopt v6
>> in a manner that is stylistically pure," or "Adopt v6 in a manner that
>> fits
>> the existing business models which make the net work," don't hold your
>> breath waiting to figure out which one I'll choose.
> It isn't a question of stylistic purity vs. fitting existing models.  It
> is more a question of peeing in the pool to make the water warmer.

Semantics. Your analogy is more colorful, but it doesn't change the basic

> Simply, PI does not scale using the existing routing technology that
> makes the net work.  The real question is pay now or pay later.  Not
> surprisingly, we'll be paying later.
> It is sad we have not learned from past mistakes.

I'll tell you a secret, I agree with you. But at the end of the day, I
don't care. Fred hit it on the head with his followup to your post, the
people that matter don't want to deploy the better solution. Period.

I could go on and on about the people and enterprises that I've talked to
who won't even consider IPv6 without PI, etc. etc.; but I don't need to.
Just look out your window and observe the current state of deployment.
That's said with all due respect to the people who actually _have_
deployed IPv6, we need all the steps forward we can get. But I am
convinced that PI space is what we've needed for some time to take
deployment to the next level.

>>> Yeah, it's nice that ISPs will eat the cost of the router upgrades.
>> Yep, just like they have since the beginning.
> I didn't think I needed the half-smiley.  Of course, ISPs never have
> eaten the costs, they pass them on or go out of business.

I think you got some collateral damage from the rounds that were aimed at
Iljitsch. I know _you_ don't need the economics lesson.

>> It's this little thing called progress ...
> No.  It is called "repeating history".  "Progress" would have been IPv6
> being designed and implemented with a scalable routing architecture.

If your definition of "scalable routing architecture" includes not tying
an enterprise (be it medium sized ISP, big company, whatever) to a single
provider for their IPv6 bandwidth, then once again, I agree with you, but
it still doesn't matter. The equation is simple. We cannot solve the
routing problems till we get experience. We can't get experience until we
get deployment at a meaningful level, and we can't get that until we get
PI space.

Twenty years from now, I strongly suspect that we'll look back on this and
realize that we were both a little right, and a little wrong. My gut tells
me that the real solution to this has yet to appear, but it will never
even be a glimmer in someone's eye until the problem actually exists to
solve in a way that matters.


    If you're never wrong, you're not trying hard enough

More information about the ipv6-ops mailing list