[quagga-dev 1071] BGP and ORIGIN (UNH IOL Test BGP-4.1.10)

Hasso Tepper hasso at estpak.ee
Thu Apr 22 07:52:40 BST 2004


My coworker Rivo Nurges tried to fix BGP-4.1.10a failure and 
discovered that things are quite messy. After some discussion he 
asked me to summarize our findings.

http://www.ietf.org/internet-drafts/draft-ietf-idr-bgp4-23.txt

<quote>
a)   ORIGIN (Type Code 1):

ORIGIN is a well-known mandatory attribute that defines the
origin of the path information.  The data octet can assume
the following values:

Value      Meaning

0          IGP - Network Layer Reachability Information
           is interior to the originating AS

1          EGP - Network Layer Reachability Information
           learned via the EGP protocol [RFC904]

2          INCOMPLETE - Network Layer Reachability
           Information learned by some other means
</quote>

EGP stuff is clear (most of routers today don't support RFC904 
anyway :).

But IGP vs. INCOMPLETE leaves enough room for interpretation and 
vendors seem to use this room with full power:

Cisco - only routes which are specified with with "network x.x.x.x/x" 
statements will have origin IGP, all other stuff (routes 
redistributed into BGP) will have origin INCOMPLETE. This is how 
Quagga bgpd behaves as well.
Logic behind this is probably - "we put IGP only to routes we know 
100% that they are interior to the our AS, we can't be so sure about 
ospf etc routes, they might be redistributed into ospf from eBGP by 
other ASBR". 

Juniper - all routes router originates into BGP, no matter is it via 
redistribution ospf, is-is, static or connected routes, will get 
origin IGP.
Logic probably - "If you are crazy enough to redistribute routes from 
IGP into eBGP, you have to really know that all this stuff is 
interior to the your AS".

UNH IOL test - as I read it, routes redistributed from ospf/is-is/rip 
etc. should get origin IGP, but redistributed static (and connected?) 
routes should get origin INCOMPLETE.

IMHO all interpretation have their own points and I can't say that any 
of them are wrong. Just BGP-4 standard isn't specific enough. But if 
all interpretations are correct (IMHO it would be nice to have some 
sentences about it in standard), UNH IOL BGP-4.1.10 test is just 
pointless.


-- 
Hasso Tepper
Elion Enterprises Ltd.
WAN administrator



More information about the Quagga-dev mailing list