From: Howard C. Berkowitz (hcb@gettcomm.com)
Date: Thu Apr 22 2004 - 00:49:07 GMT-3
At 9:22 PM -0500 4/21/04, Kaiser Anwar wrote:
> I will definitely read the book. But just wondering under what
>circumstances would want to get the whole routing table or become transit
>AS. Thanks
>
You are mixing two distinct concepts. Whether or not you take full
routes [1] s separate from the policy of being a transit or end AS.
You might want to look at a terminology document of which I'm one of
the authors, and has been approved for RFC after we make a few
editorial changes, such as [1] below.
http://www.ietf.org/internet-drafts/draft-ietf-bmwg-conterm-05.txt
Assume, for example, that Cisco doesn't provide Internet access to
other than its own offices. In reality, there are special cases for
contractors and the like. Given Cisco connects to ISPs all over the
world, and has extensive internal connectivity, and they get AWFULLY
good prices on big Cisco routers, it would be wise for them to take
full routes at all their major campuses, for traffic engineering and
fault tolerance that considers a wide range of exit points. This is
an extremely large case of taking full routes at multiple points, but
still being essentially an end AS.
A small ISP, however, might buy transit from a national provider,
perhaps multihoming to the same ISP at different POPs. The small ISP
might simply accept default to the upstream provider, and advertise
only its customer routes to the upstream. This is an example of
being a transit provider with a very small routing table.
[1] It's worth remembering that there is no such thing as a master global
routing table. Even when you take full routes, it is one view that depends
on your peering with other AS. You might never see certain routes because
their upstream doesn't propagate.
This archive was generated by hypermail 2.1.4 : Mon May 03 2004 - 19:48:52 GMT-3