>- > > From: John A. Kilpatrick [mailto:[EMAIL PROTECTED] >> Sent: Wednesday, May 14, 2003 5:35 AM >> To: [EMAIL PROTECTED] >> Subject: Cisco and MED [7:69060] >> >> >> One question I have....why does Cisco refer to MED as >> "metric" all the time? >> I admittedly am new to BGP on Cisco (I've done it a bit on >> Junipers) but >> when you do a `show ip bgp` one of the columns is "metric", >> which appears to >> be MED, and when you do a route-map where you `set metric >> blah` it seems to > > set MED. So in Cisco's use of BGP do they just call MED the "metric"?
It's really not so much Cisco versus Juniper as BGP-3 versus BGP-4, and later experience with BGP-4. In BGP-3, the equivalent attribute was called "Inter-AS Metric." The semantics were a little less worked out than in the first BGP-4 specification. In general, the assumption was that it was an IGP metric redistributed into BGP. When BGP-4 came out, the semantics were tightened so the attribute was for use with directly connected AS only (i.e., non-transitive) and with the restricted purpose of selecting among different exit points to the same AS (i.e., multi-exit discriminator). With industry experience, however, there came to be recognized there were other applications for an adjacent-AS-only attribute, such as selecting among several directly peered AS at a multilateral exchange point. Hence, the knob to compare MED among several AS, with the constraint they all be adjacent, was defined. Avi Freedman did an excellent operational tutorial on this at the Denver NANOG -- I did the basic BGP tutorial and he did the intermediate. I should mention that the IETF clarified, in some of the many, many drafts of the emerging RFC 1771 updates, some things that are more complex than a simple redistributed IGP metric -- how to treat the case of a missing MED: best or worst? This resulted in another knob, since Cisco's default is the opposite of what the IETF finally decided later. Things like BGP-deterministic-MED came into being to clarify more special cases of MED, and especially IGP metric redistribution, in reflectors and clusters. See RFC 3345. Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=71810&t=69060 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
