[ppml] Policy Proposal: IPv4 Soft Landing

[Rich Emmings]

On Wed, 16 May 2007, David Conrad wrote:

> Hmm.  I'm curious what you see as the idea that was poorly conceived in 
> 2007-12 and weren't addressed in Soft Landing.  I considered Soft Landing to 
> be pretty much the diametric opposite of 2007-12 (no reservation, no freezing 
> of policy).

Creation of artificially early dates for restriction to resources will 
increase demand for those resources earlier.  One example that comes to 
mind is US 1970's gas distribution issues.

>> 1) Based on conversations, I get the impression that most folks proposing
>> implementation IPv6, _probably_ have not tried to implement IPv6 on any
>> large scale.  ping6 -I eth0 fe80::xxxxxxxxxxxxxxxx doesn't count.  (I will
>> grant I could be wrong on this, hence "probably".)
>
> Guilty as charged.  I have not implemented IPv6 on a large scale.  However, 
> neither has most of the world, and part of the rationale behind Soft Landing 
> is to encourage people to start.

I think the existing ARIN policies have done a great job at encouraging those who can start, to start, but you can't push a rope:

Full support in some routing platforms only just got here.  Tools like ndp (think arp) are not present, or not easily present on hosts.

Not all ISP's support IPv6 yet.  Changing ISP's can be a problem unless you do it often.

Not all other "hardware" is ready.  Drop your average firewall into the mix, and 
it starts to get ugly real fast.  SCADA is another area where we see deficiency, 
and worse, it's a growing area.  Remote monitoring means RFC1918 is not 
always an option.

People keep trying to rewrite the IPv6 RFC's, and while some of those don't 
have a real operational impact, it makes it easy for people to say "Don't 
implement, they're still defining it.  We don't want a bloody nose for being 
an early adapter."  Makes it harder to get support to go forward.

Like it or not, Windows is a poplar platform, but it's only in Vista that 
they _support_ IPv6.  It's in other versions, but not supported.

(If someone wants to start a thread on barriers to entry to IPv6 and 
solutions to get over, around or though those barriers, it might informative 
to hear from those that are further along.)

I've been pushing for starting IPv6 for lot of years now, because I think 
it's going to take at least 5 years to go from "start" to "supportable" so 
starting early makes sense.  But implementation has been all working around 
problems and bloody noses.


>> 2) Artifically making a commodity rare, will cause a run on it before it's 
>> rare and push up the dates.
>
> Fortunately(?), IPv4 space scarcity isn't artificial.
>

But pushing up the date when we limit assignments is artificial exhaustion. 
Call it hoarding, or call rationing, but it's artificial.

IP addresses are only useful when used. The last block of IPv4 addresses 
serves no function when it's sitting on the sitting on the shelf.

What will happen when the last IPv4 address is used?  The Internet will still be there.
What will happen is speculation, with some guesses more accurate than others.


> I'm not sure I'm following you.  These sound like different policies that 
> could be applied within the framework of Soft Landing.

If thing are this critical, then we need to change the allocation policies 
now, not when we reach an artificial date.  If things are not critical, the 
existing polices are ok.

> The problem with this is that it leaves people in the dark about what is 
> going to happen in the future.  I am trying to give people a roadmap so that 
> they're aware of what is in store for them.
>

In the long run, we're all 6 feet under, and we'll run out of IP addresses 
too (IPv4 and IPv6).  So we're looking at medium term solutions.  We do a 
full roadmap, we're trying to predict the future.  Do that anywhere reliability,
and you'll make far more cash in the stock market than in data 
processing.

"This is what will happen next" still provides some information about the 
future without expending a lot of time roadmapping a future that may not 
happen.  Behavior under that system allows experience to feed back into the 
next step.

We may as well be having the conversation about Social Security, a number of 
the issues and proposed solutions are not analogous, but similar at a basic 
level, and potentially just as contentious.