[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


Pekka Savola wrote:
> On Fri, 12 Nov 2004, Bob Hinden wrote:
> >> So, are you requiring that BGP will have to have a 
> configuration knob 
> >> to to special-case fc00::/7? (This cannot be done using 
> the typical 
> >> route-map/policy statements, because it has to be on by 
> default and 
> >> not interfere with the operators configuring their own policy.)
> >
> > Not exactly.  The intent is to suggest that border routers be 
> > configured to do this.  We are not requiring knobs or special case 
> > code for this.
> This is slightly different than what Tony is proposing, but 
> let me state the problems I see with both:
>   1) operator configuration: many operators probably won't do 
> that unless they have set up filters (many have, from 6bone 
> times) to weed out unexpected prefixes.
>   2) vendor code: the only way to implement this in the sane 
> manner would be having a 'allow-ula' -like knob for BGP.  You 
> cannot configure this in the policies, because either a) the 
> operator forgets about ULAs when writing the policy, or b) 
> the policy does not exist and it isn't applied.
> So both would appear to have their problems, but 1) might be 
> slightly better, especially if GROW WG would produce a BCP or 
> some other document giving guidance on filtering.

Think in terms of the clueless operator that doesn't read BCP's or even know
they exist. My initial thought was a defaultl policy of FC00::/7 ->
/dev/null. This would allow any longer prefix to override, but would also
require explicit action to do so. Thinking in terms of BGP, the language of
accept/announce makes more sense, but as you note would require a knob to
allow any explicit action to override it. As I said, I am sure the text
needs work, but there appears to be a growing concern that 'just leave it to
the operators' is not going to work.


IETF IPv6 working group mailing list
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6