[c-nsp] route maps and bgp

[Gert Doering]

Hi,

On Wed, Oct 05, 2005 at 08:24:56AM -0500, Pete Templin wrote:
> > On Tue, Oct 04, 2005 at 10:53:26AM +0200, Alexander Koch wrote:
> > 
> >>Be careful! Whenever you have BGP customers change that
> >>please. A prefix-list is no good in that case, as when your
> >>customer is not sending you his network range you will send
> >>whatever is best in your routing table. Usually you then
> >>'leak' that range from your transit (because it's best).
> >>
> >>BGP communities is the only sane thing to do here.
> > 
> > I tend to disagree.  For a network that gets such basic things wrong, a
> > configuration that errs on the side of *not* announcing things is 
> > much better for everyone involved.
> 
> There might have been a misunderstanding here - what Alexander wrote is 
> correct; if the customer stops announcing the prefix to <you> but you 
> are still learning the prefix from <other-provider>, the prefix will 
> match the prefix-list and <you> will provide transit from 
> <other-provider> to <provider>.  

OK, you're right here.  I was thinking specifically about the setup in
question, where both prefixes are generated with local "network" 
statements, and no customer network blocks involved.

What we've been doing "at that time in BGP development" was to add an
outgoing as-path access-list as well, that had an

   permit ^$
   deny .*

to make really, *really* sure we're not going to announce 3rd party ASes
to our upstreams...

gert

-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert at greenie.muc.de
fax: +49-89-35655025                        gert at net.informatik.tu-muenchen.de