Sean puts this very nicely... I was away today so I missed the rest of the
traffic and looking it over alot of it was not relevant. I'll put in some
comments here though.
> What about the other large isps? What would it take for you to do
> something? Chris is gracious enough to show up and participate, at
> least even if it does mean he has to wear nomex.
I'm in favor of source address filtering at the edges.
Sure, so are most operators, including me. Define the edge the right way,
as Sean says, though. The edge is the customer CPE, or even the LAN router
inside the customer network. As close as possible to actual hosts that
can spoof.
I'm opposed to some of the suggestions where to put source address
filters, especially placing them in "non-edge" locations. E.g. requiring
address filters at US border crossings is a *bad* idea, worthy of an
official visit from the bad idea fairy.
In general, knowing what ips a 'peer' is going to send you is very
difficult. One would hope that all peers would have all customers
implement filtering as close to their hosts as possible. This isn't a
reality today 
Putting filtering, of any type, in a 'core' spot is asking for trouble.
Doing much other than routing at the core is a bad plan. In the future
perhaps filtering and other magic would be possible in the core, but in
reality it will almost always be a tough thing to do.
Our networks, and our customer bases, are not identical. This is good
and bad. Not to make excuses, the ease with which this can technically be
done depends on your network and type of customer connections. Or more
precisely how you aggregate customer connections.
This is a good point. If you are aggregating customers on gear where the
uRPF filtering is done in software you will severely impact performance
with anything over oc3 rates
Do not confuse acls with uRPF, acls on t1 aggragation isn't workable
unless your 'network' lives in your basement. uRPF is only feasible in
some small circumstances.
Making the assumption that because your network works with uRPF and thus
Sean's or Ryan's or Paul's will isn't really valid. I would like to
believe that the majority of NSP/ISP operators have atleast considered
uRPF, or some kind of mitigation techniques. We have, and for us the
current possibilities aren't feasible for a number of technical reasons
which the respective's vendors are aware.
IMHO the edge is generally the best place to do source address filtering.
After traffic is aggregated its more difficult. Some folks have already
identified the technical limitations of some types equipment. And with
the market, we're going to be stuck with that equipment for a while. In
hindsight, if every provider and every equipment vendor did it from day 1,
we would be in great shape.
Yes, filter at the 'edge', where 'edge' == 'as close to the host as
possible'. Possibly this could be accomplished by a config default in
IOS/JunOS... something akin to 'no ip directed broadcast'... or atleast
that possibility was raised at the NANOG Security BOF.
This won't solve any problems in the short term, but it is a move in the
right direction.
Does that mean I can wave my magic wand and fix everything tomorrow? No.
Can I work on standards, vendors and purchasing agents to change this over
time? Yes. Will yelling at me make it happen faster? I doubt it, but I
know you will anyway.
This is the tact that we are taking inside UUNET today. We are trying to
work on a consensus document for 'Network Equipment Requirements'. This
effort is being headed up by Mr. George Jones, I believe this document is
headed out for IETF work and possibly additions to BCP 38 ? Barry Greene
is helping out on this along with some folks from Merit.
The focus of this is to get everyone possible asking for the same thing
from their vendors. For a long time every vendor through the door would
say: "security what?", "Filtering what?", "You are the ONLY person asking
for this capability." This was the response to: Radius authentication for
management access, tacacs auth for management access, auth for management
access, filtering, filtering at line rate, filtering on all interfaces...
the list goes on of things that one would assume are 'standard' on any
network gear. UUNET, atleast, has been asking all vendors we
see/speak-with/test for the list of things in George's document for the
last 2 years.
-Chris "Nomex, whats that?" Morrow.