Same AS number from different location and Migration of IPaddresses

Sat May 24 20:22:17 UTC 2008

Patrick,

Your usage is quite consistent with the RFC 1930 guidelines on the use of
AS, which probably does need some updating but does have an operational
rather than a protocol theory viewpoint. 

Specifically, an AS is defined not as a business entity, not as a routing
domain, but as:

   "...a connected group of one or more IP prefixes run by one
       or more network operators which has a SINGLE and CLEARLY DEFINED
       routing policy."

In this case, the sites have a common, coordinated routing policy. I do
agree that practicality does call for them to have a direct connection, but
otherwise, they meet the requirement of being one or more IP prefixes run by
one or more operators.

I do hope they register their routing policy, with appropriate comments.

Howard

-----Original Message-----
From: Patrick W. Gilmore [mailto:patrick at ianai.net] 
Sent: Saturday, May 24, 2008 11:11 AM
To: NANOG list
Subject: Re: Same AS number from different location and Migration of
IPaddresses

On May 24, 2008, at 9:15 AM, Marshall Eubanks wrote:
> On May 23, 2008, at 8:15 PM, devang patel wrote:
>
>> Is that okay to use Same AS number for the two different site on  
>> different
>> location?
>
> To answer this specific question, Autonomous Systems should be  
> topologically convex.
> This means, at the Internet interdomain routing (BGP) level, that  
> packets
> cannot leave an AS in one place to get to locations in the same AS  
> in some other place.
>
> So, to put two sites on one AS, there should be an internal  
> connection between them, which can be done
> through your internal network, by a direct connection, or by a  
> tunnel. Traffic might come to
> the AS at either site, and has to be routed internally to get to the  
> other.

I am afraid I have to disagree with Marshall.

The idea behind an AS when the routing protocols were written long ago  
may have been a contiguous domain, but there are lots of things the  
protocols did not originally envision.

If you have two islands, and they each have a prefix which is globally  
routable, there is nothing wrong with the two islands sharing a single  
ASN.  Island A announces Prefix A, and Island B announces Prefix B.   
Routing is done by prefix, not ASN, so there is no fear of Island A  
getting packets for Island B, and therefore no requirement for  
internal connectivity.  And before anyone says anything about Island A  
not having connectivity to Island B, these are obviously not "transit  
free" networks, so each island can just point default.   In fact,  
cisco even has a knob to listen to paths with your own ASN in it so  
you can do this without default (although I'm not sure I'd recommend  
that).

It works fine and saves the community from burning an ASN.

-- 
TTFN,
patrick