Re: Best Common Practice - Listening to local routes from peers?

2004-02-27 Thread Stephen J. Wilcox


On Thu, 26 Feb 2004, Michael Smith wrote:

 We have a customer of a customer who is attempting to send traffic from
 IP space we control, through the Internet and back into us via one of
 our transit connections.
 
 I have filters in place that block all inbound traffic from the blocks I
 announce coming in over my transit and peering connections.  This is
 breaking the downstream customer ability to route from them, through
 UUNet, and back to me.

Yes, I've had this back in the days when I  used to attempt to do fascist 
filtering and security ... the short answer is you cant do this kind of 
filtering in the backbone, you need to push it to the edge (defined in my mind 
as stub areas of network.. in this case thats likely not in your network but in 
the customer's network)
 
 I'm curious what the Best Common Practice is for this type of scenario. I have
 always used this type of filtering as a way to bury source-spoofed traffic in
 a DDOS situation but I'm not sure if it's appropriate, generally speaking.

Am not convinced the benefit of dropping that traffic is worth the effort tbh 
(that is stuff coming in with obviuosly spoofed addresses.. there is so much 
legit space available to spoof).

Steve

 
 If other operators would like to reply directly to me I would be more
 than happy to summarize to the list.  Thank you for any assistance you
 can provide.
 
 Michael Smith
 [EMAIL PROTECTED]
 
 




Re: Best Common Practice - Listening to local routes from peers?

2004-02-27 Thread Patrick W . Gilmore
On Feb 27, 2004, at 1:21 AM, James Edwards wrote:

On Thu, 2004-02-26 at 21:22, Michael Smith wrote:
Hello:

We have a customer of a customer who is attempting to send traffic 
from
IP space we control, through the Internet and back into us via one of
our transit connections.

Gotta ask, Why ? They have a direct connection (few hops) to you and 
are
trying to go the long way, right ?
Well, there are possible cost considerations.

Also, and perhaps more importantly, if the customer's line to his 
network drops, should the customer be incapable of getting to content 
hosted on his network?

--
TTFN,
patrick


Re: Best Common Practice - Listening to local routes from peers?

2004-02-27 Thread james


: Also, and perhaps more importantly, if the customer's line to his
: network drops, should the customer be incapable of getting to content
: hosted on his network?


Simple. I am not going to break something all the time to counter something that might 
break every once and a while.
When Nagios pages me that a multihomed customer is down, I will allow that address 
space to enter all my gateways, as a source
address.



James Edwards
Routing and Security Administrator
[EMAIL PROTECTED]
At the Santa Fe Office: Internet at Cyber Mesa
Store hours: 9-6 Monday through Friday
505-988-9200 SIP:1(747)669-1965



Re: Best Common Practice - Listening to local routes from peers?

2004-02-27 Thread Patrick W . Gilmore
On Feb 27, 2004, at 1:51 PM, james wrote:

: Also, and perhaps more importantly, if the customer's line to his
: network drops, should the customer be incapable of getting to content
: hosted on his network?

Simple. I am not going to break something all the time to counter 
something that might break every once and a while.
When Nagios pages me that a multihomed customer is down, I will allow 
that address space to enter all my gateways, as a source
address.
To quote someone on this list: It doesn't scale.

Additionally, if I'm a customer and my line goes down and it takes you 
15 minutes to fix it, I'm not going to be happy about it.

Finally, you fix it at the *edge*, that won't break either.

--
TTFN,
patrick


Re: Best Common Practice - Listening to local routes from peers?

2004-02-26 Thread Patrick W . Gilmore
On Feb 26, 2004, at 11:22 PM, Michael Smith wrote:

We have a customer of a customer who is attempting to send traffic from
IP space we control, through the Internet and back into us via one of
our transit connections.
I have filters in place that block all inbound traffic from the blocks 
I
announce coming in over my transit and peering connections.  This is
breaking the downstream customer ability to route from them, through
UUNet, and back to me.

I'm curious what the Best Common Practice is for this type of scenario.
I have always used this type of filtering as a way to bury
source-spoofed traffic in a DDOS situation but I'm not sure if it's
appropriate, generally speaking.
It is a good idea to filter source IP on the edge.  Since your customer 
has more than one upstream, filtering their IP space at your border is 
not the edge.

Filter their source IP where your network meets their network.  Filter 
your source IP at your upstream borders.

My $0.003411284. :)

--
TTFN,
patrick