[ppml] v6 multihoming and route filters

[Iljitsch van Beijnum]

On 30-jun-2006, at 9:40, Fred Baker wrote:

> My opinion - and please note that it is just that, not an edict of  
> any kinds - is that in the final analysis it is not the IETF but  
> operational reality that controls the issues here.

There are two issues with having people inject /48s into the routing  
system without limitations:

1. At some point in the future, the number of routes in the DFZ could  
become so large that the routing system can't support it any more.  
This is the scaling problem that many people in the IETF are worried  
about.

2. A more operational and more immediate risk is leaking of more  
specifics. Due to the way they handle their internal and external  
routing, and the relationship between the two, it's not uncommon for  
larger networks to leak internal more specific routes into BGP. This  
way, an ISP with a /16 with a number of customers that each have a / 
24 may leak those /24s. In IPv6 this is very dangerous, because a  
single ISP with a /32 can have a maximum of no less than 65536 /48s  
that can potentially be deaggregated. So one ISP could potentially  
leak a number of routes equal to a third of the global IPv4 routing  
table.

The second issue could largely be solved by giving multihomed  
customers a prefix that is shorter than a /48 and then filter out / 
48s but allow /47 and shorter. But that's not the way things work today.

> What may have applicability is Steve Deering's concept of  
> Metropolitan Addressing, which it looks like someone needs to  
> describe in an Internet Draft (it is in the slides at ftp:// 
> ftp.ietf.cnri.reston.va.us/ietf-online-proceedings/95jul/ 
> presentations/allocation/deering.slides.ps).

I've written a draft that takes this idea a bit further a couple of  
years ago, have a look at:

http://www.muada.com/drafts/draft-van-beijnum-multi6-isp-int-aggr-01.txt