[c-nsp] BGP network stops being advertized

[Mark Tinka]

On Tuesday 10 June 2008, Deepak Jain wrote:

> In the "old" days, null was handled by CPU (software
> switched), so lots of us old-timers got into the habit of
> using loopback instead of null. On a modern platform it
> should make no operational difference provided you have
> everything you need set up properly. (null routing the
> /16, for example, might be bad if you ever actually try
> to route that whole prefix instead of just subs -- as it
> would be specific and static).

How often is this the case, though?

If a /16 were routed to a downstream customer, then it'd 
point to an interface and/or remote IP address.

One could "subnet" a /16 on an interface and assign 
addresses out of that with a 16-bit long netmask, but aside 
from NAT'ed environments, I haven't quite seen this in 
production.

We null route our aggregates (/19's, /18's, e.t.c.), and 
originate them on our route reflectors for onward 
announcement to customers, upstreams and peers. We tag them 
with a next-hop that sends them to the Null0 interface on 
edge/peering routers, should any of these destinations 
attempt to send traffic to directly to them.

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 832 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20080610/0da52e9a/attachment.bin>