[ppml] alternative realities (was PIv6 for legacy holders (/wRSA + efficient use))

[David Conrad]

John,

On Jul 31, 2007, at 8:38 PM, John Curran wrote:
> The burn rate is between 10 to 15 /8's per annum, increasing,
> and while you'll see some pieces of the legacy space show up,
> the extractable and reusable space will burn off under that
> demand in a few short years.

So you do not believe that under pricing pressure, people will be  
more efficient in address space usage thereby reducing the  
consumption rate?  This seems counter intuitive to me.  I once had 6  
static IP addresses for my home.  I reverted to one dynamic IP  
address + NAT exclusively because it was cheaper.  Perhaps I'm overly  
price sensitive.

Further, I would assume the largest address consumers, faced with non- 
predictable costs in obtaining resources will likely accelerate  
deployment plans for IPv6 for their internal infrastructure where  
possible, thereby freeing up the IPv4 addresses they're using for  
that infrastructure for customer assignment further reducing the  
consumption of the "available" pool.

Seems to me that the demand for IPv4 will only decrease over time  
(although it may spike during a land rush prior to free pool  
exhaustion, depending on how the policies evolve).

> very quickly the largest ISP's
> will face having to ignore new routes from their peers (or they'll
> be seeing if they can replace every EGP router to some of "2M"
> route variety, only to do it again every ninety days...)

I'm curious how you derived 'every ninety days'.  I would assume that  
if a market were to develop, shorter prefixes would be more desirable  
than longer and ISPs would make deals like "free service for life  
plus this new car if you cede your /16 to me" or some such.  Yes, as  
time goes on, there will be a trend towards longer prefixes as the  
aggregatable prefixes get consumed, but there will _always_ be the  
back pressure of the prefix length filters as they exist today.

> Markets
> aren't hierarchical, and there's no working backpressure model
> for the imputed non-hierarchical routing cost,

Sure there is.  It's called prefix length filters -- that's what  
worked the last time we were in this situation. As I've said before,  
I wonder who is going to be playing the part of Sean Doran.  The fear  
of obtaining unroutable prefixes means there is a limit to the length  
of prefix that will be on the market.

> so it will spin apart sooner or later.

"In the long term, we're all dead." -- John Maynard Keynes

Rgds,
-drc