Customer IPv6 range assignments.

[Joe Abley]

On 26-Jul-2006, at 15:40, Stephen Fulton wrote:

> According to the ARIN documentation we've read, the standard  
> assignment to end-users (whether a single person or a large  
> corporation) should be a /48.  Regardless of the amount of address  
> space available with IPv6, this seems like an awful waste of  
> space.  I'm curious if this policy is still current, or have I mis- 
> interpreted the documentation?  Would we be breaking rules if we  
> assigned a /64 or /56 to a small client?

I seem to think you can assign a /64 in the case that you're sure  
that only one subnet will ever be required, and a /128 in the case  
that you're sure that precisely one interface address will ever be  
required. However, I haven't read the v6 assignment policies at ARIN  
for a year or so, and things may have changed.

When you're used to justifying the need for address space with IPv4,  
the idea of assigning a /48 to a user who might well only need to  
number one subnet seems ridiculous. However, when you further  
consider how much address space there is in IPv6, it doesn't look (to  
me) like a dangerous strategy.

 From many angles, fixed-length /48 assignment to end users is  
classful addressing. Although it can seem initially horrifying to say  
so, classful addressing had a lot going for it.

The principle thing wrong with classful addressing in IPv4 was that  
it resulted in assignment waste which was not sustainable given the  
number of IPv4 addresses available in a 32-bit word, and the rapid  
growth in end users.

With 79228162514264337593543950336 as much address space available in  
IPv6 compared to IPv4, my view is that the convenience and simplicity  
associated with classful addressing far outweighs the danger of  
address exhaustion.

Others have suggested that some boundary between 48 and 64 could be  
chosen which would facilitate fixed-length assignments to customers  
without incurring so much waste. I agree with those people, but I  
don't see the reason for making the change given that the 48-bit  
boundary has some history and has seen some deployment.

The mileage of others definitely varies.


Joe