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


To: Patrik Fältström <paf@cisco.com>
cc: dnsop@cafax.se
From: Bruce Campbell <bruce.campbell@ripe.net>
Date: Mon, 4 Nov 2002 19:58:04 +0100 (CET)
In-Reply-To: <7588BEEC-F00C-11D6-868E-0003934B2128@cisco.com>
Sender: owner-dnsop@cafax.se
Subject: Re: DoS and anycast

On Mon, 4 Nov 2002, [ISO-8859-1] Patrik Fältström wrote:

> On måndag, nov 4, 2002, at 16:28 Europe/Stockholm, Randy Bush wrote:
>
> > second, as many of us have repeatedly said, routing of anycast
> > addresses has to be appropriately scoped, as it has to be today.
> > it would be useful to have a discussion of 'appropriately' if we
> > could stop ratholing on other issues.
>
> I thought I had a very specific question?
>
> Default route for ISP A is to B, for B is to C. If C _internally_ have
> a copy of the IP address 1.2.3.4 for it's own use, will traffic from B
> to C reach that server, or the real 1.2.3.4 which is at D which B ask C
> to transit traffic to?

Ok.  Assuming that C's transit routers listen to C's internal routing,
then traffic to 1.2.3.4 from A will be directed to C:1.2.3.4, not
D:1.2.3.4 .  This is probably the most-likely case.

If C have Clue++, and have directed their transit routers to _not_ listen
to that specific internal announcement, then traffic from outside of C
will go (as non-C customers expect) to D:1.2.3.4 .

Traffic from C's customers (etc) will always go to C:1.2.3.4 .

--==--
Bruce.


#----------------------------------------------------------------------
# To unsubscribe, send a message to <dnsop-request@cafax.se>.

Home | Date list | Subject list