[arin-ppml] IPv6 /32 minimum for extra-small ISP

[Owen DeLong]

On Apr 14, 2010, at 12:58 PM, Milton L Mueller wrote:

>> -----Original Message-----
>> The problem is that even if you got an extremely small allocation
>> that as soon as you advertise it, you consume the same amount of 
>> routing resources on everyone else's BGP router on the Internet, 
>> that someone with a larger allocation has who is advertising theirs. 
>>  Thus it is not fair to base pricing solely on the amount of numbering.
> 
> Just curious. Has this assertion ever been tested, examined, modeled?
> There are two logical components to this statement that I am interested in verifying: 
> A small allocation: 
> a) "consumes the same amount of resources" 
> b) on "everyone else's" BGP router
> 
Milton,
	A globally visible prefix is a globally visible prefix.

	If your prefix is globally visible, then, yes, it takes a single
prefix slot in everyone else's "full-table" router(s).

	There is no difference in the amount of table space occupied
on any router between a /33, a /48, and a /128 unless one of them
is de-aggregated into multiple prefixes.

These facts are routing 101 nature of how routers work. They are
not logical assumptions that need to be modeled. They are simple
hard fact of how the system is designed and operates.

Owen

(Note, I still think that PI for whoever needs it is a good idea and that
the routing system will find a way to adapt well within the required
timeframe. As such, if it were somehow possible to prove these
facts wrong, I would have a vested interest in doing so, but, even
I recognize that these are the basic facts of the situation.)