[ppml] Another question on Policy Proposal 2007-8

>-----Original Message-----
>From: ppml-bounces at arin.net [mailto:ppml-bounces at arin.net]On Behalf Of
>Kevin Loch
>Sent: Thursday, March 08, 2007 12:20 PM
>To: ppml at arin.net
>Subject: Re: [ppml] Another question on Policy Proposal 2007-8
>
>
>Ted Mittelstaedt wrote:
>
>> Lest you think that upstream networks are blocking, think again.  We
>> acquired an ISP last year connected to cogentco.  I was rather amused
>> to find that I could advertise any number block I wanted through
>> cogentco, they do -no- BGP route filtering of any kind, at least not
>> that I could find.
>
>Did you have bgp customers yourself?

That isn't the issue.  You can have downstream bgp customers just fine but
if neither they or you are providing transit to the general Internet then
IMHO your feeds are doing a disservice to the Internet by not
filtering.  We aren't talking general traffic here.  We are talking route
advertisements.  (and by the way, yes we do have 1 downstream bgp customer
with his own numbering)

>If your upstreams don't generate
>filters automatically from rr entries and you do have enough downstream
>bgp customers this is not that unusual.

It is normal to NOT filter bgp advertisements when your peering or your
have customers that are peering.  I understand this and that isn't the
problem.

The problem is bozo-brained admins on peering networks who have end-node
AS customers who all have known, defined AS's that they are advertising
(or forwarding advertisements for) and they are allowing them to inject
any advertisement they want, without any filtering.

>Some may trust that you are
>filtering all of your downstream customers and some may just not care
>until you start screwing up.
>

There really is no justification for not filtering BGP advertisements from
non-transit, end-node ASs.  We have had enough problems in the past with
this.
I recall 3-4 years ago Sprint took half the Internet offline by advertising
0.0.0.0 for a good 4-6 hours.

Ted