Marco Marzetti wrote: > Beside i agree on the most part of this draft i would like to underline > how bullet 2.2.2 breaks the the nature of BGP itself. > > The exact part i am referring to is: > """ > As a route server does not participate in the process of forwarding > data between client routers, and because modification of the AS_PATH > attribute could affect route server client BGP Decision Process, the > route server SHOULD NOT prepend its own AS number to the AS_PATH > segment nor modify the AS_PATH segment in any other way. > """ > > I firmly think that it really has to state "MUST NOT" instead of "SHOULD > NOT". > > From a route-server client point of view that breaks the natural BGP > best selection process and forces you to purely rely on LOCAL_PREFERENCE > which in turns breaks your traffic engineering because it forces you to > prefer/not-prefer the prefixes learned over the route server in place of > those learned, for instance, over a private session or another IXP. It doesn't "break" the decision process mechanism. Localpref might be ugly, but using it doesn't violate the bgp protocol. This statement needs to be "MUST NOT". Technically you can run a route server even if you insert the intermediate ASN into the path. The only absolute "MUST" is that you mustn't change the next-hop; otherwise you end up with a router instead of a route server. RFC2119 is a pain because there is no way of differentiating between "SHOULD NOT", meaning "on balance it would be a better idea not to do X" and "SHOULD NOT", meaning "if you even _think_ about doing X, you need your head examined and the only reason we specified SHOULD NOT instead of MUST NOT in this situation because of some technicality that prevented us from doing so". The text in 2.2.2 veers slightly in the direction of the latter interpretation. Nick