Re: v6 multihoming and route filters

[Fred Baker <fred@cisco.com>]

On Jul 4, 2006, at 7:04 PM, Marc Blanchet wrote:

> the fact that I'm suggesting to filter at the /48 level is just  
> that I don't know how to build a concensus right now on another  
> more restrictive boundary.  for the global routing table  
> perspective, I would love to recommend /32 or smaller, but I'm not  
> sure we can get that agreement.
I'm not in the middle of ARIN etc, and frankly think they know their  
business better than I know their business. Consider this my two yen  
from a distant part of the peanut gallery.
That said, the counterpart here is to say that a typical business  
gets a /28 IPv4 prefix (or whatever it gets), so the obvious BGP  
filter level is /28 in IP. That's no a meaningful recommendation,  
even if it was acceptable - you're never going to get anything  
smaller than that in the first place.
In the IPv4 world, the guideline is generally that one filters to the  
allocation level (/20 or /19 or whatever - it changes) and goes for a  
longer one only if one has a business reason to do so. AFAIK, the  
counterpart in IPv6 is a /32. So it seems like the initial  
recommendation for IPv6 should be as it has indeed been - that one  
filter to the allocation level unless one has a business reason to  
accept a longer one.
If the operational community wants to split the difference somewhere,  
so be it. I could imagine a /40 if the operational community decided  
that scaled adequately for their needs. But if you're filtering at / 
48, there is no point in filtering.