[ppml] IPv6>>32

[David Conrad]

[Speaking only for myself]

Tony,

On May 11, 2005, at 3:12 PM, Tony Hain wrote:
> You appear to be thinking in terms of existing fixed networks and the
> cumbersome tools that vendors provide for managing those. It is not  
> hard to
> imagine a personal area and/or vehicle network where the handset/ 
> router
> changes providers between available short range and long range  
> radios over
> fairly short time periods.

It is also not hard to imagine world peace, the hard part is actually  
getting there...

> Renumbering is not a complex task.

If renumbering was not a complex task, then no one would need worry  
about provider independent space.  In a note you sent to PPML on  
4/27, you said (when discussing a non-provider-based routing approach):

>> ] Given a
>> ] choice between provider-lock/renumbering-pain vs. a little bit  
>> more for a
>> ] diverted fiber run, I am sure businesses would favor the known  
>> cost of fiber
>> ] over the unknown and potentially lethal costs of yielding to the  
>> provider.

I don't think you can have it both ways.  If renumbering is easy, no  
one would worry about the cost of renumbering and there would be no  
provider lock.  If renumbering isn't easy, people with significant  
network infrastructure would be (rightly) worried about the cost of  
renumbering and hence be susceptible to being locked into a provider.

My contention, which has at least been corroborated by a couple of  
people from larger enterprises, is that renumbering actually is a  
complex task and will remain so for the foreseeable future given the  
current IPv6 architecture.  For this reason, if IPv6 is going to be  
deployed in significant amounts, we will be exactly repeating the  
history of IPv4 address allocations with the folks getting /19s and / 
20s today being the equivalent of folks getting class As in pre- 
history.  Assertions about the huge number of /48s or /45s are  
irrelevant given medium size telcos are getting /19s and /20s.  As  
such, it would seem to be prudent to be somewhat conservative in  
allocations.

> Comparing IPv6 allocation practice to IPv4
> practice is not overly useful because one is allocating the demarc  
> for a
> network while the other is allocating for a specific number of  
> devices.

Interesting observation.  This would appear to argue that the entire  
concept of the HD ratio is spurious since host density would be  
irrelevant.  Hmm.

Rgds,
-drc
[Speaking only for myself.  Really]