[ppml] Policy Proposal: IPv6 Assignment Guidelines

[Scott Leibrand]

Iljitsch van Beijnum wrote:
> On 23-aug-2007, at 1:31, Scott Leibrand wrote:
>
>> I've heard this argument (in favor of /60 instead of /56) before, and 
>> either would be fine with me, but I still don't understand why 
>> renumbering from a PA /56 into a PA /48 would be any harder than 
>> renumbering from a PA /48 to another provider's PA /48.
>
> The choice isn't the one between renumbering from a /56 to a /48 vs a 
> /48 to a /48, but from a /60 to a /48 vs a /56 to a /48. I think /60 
> -> /48 is a lot less painful. Once you move from a /60 to a /48 (or 
> skip the /60 because you know you're not a residential user in the 
> first place) you won't have to renumber when you deploy more subnets; 
> only when changing ISPs.

Ok.  I'm assuming that networks would renumber due to change ISPs a lot 
more often than they would need to renumber due to network growth.  I 
would also anticipate that any network growing larger than a /56 would 
qualify for an IPv6 PI /48, and would therefore only need to renumber 
once.  I don't think the same could be said of all networks growing 
beyond a /60.


-Scott