> Community: > 1. Good point to remove own communities, since they have local > significance only. > 2. Rest of the communities should be preserved.
Overloading the community path has been a fuzzy attack vector that has crashed BGP devices in the past. I would be surprised if the vendors kept this in their regression testing. I think allowing long AS communities with zero control is some that needs to be looked at from all the angles. > > Bogons: > Agree that all kind of bogon, long as path, filtering can be realized in > both directions. Filtering in both direction is the Murphy filter (Myrphy's law). Many operators have been caught thinking that "I'm never going to leak" only to have a surprise in the middle of the night. That is why in my "BCP materials" I've always taught to put the BGP policy filters in both directions. > Prefix-length: > As a common recommendation I think both ISPs and non ISPs should consider > /24(ipv4) and /48(ipv6) as longest prefix allowed to be routed over the > Internet. Since this is kind of common agreement through Internet community, > BGP policy implementation recommendations should consider only those prefix > lengths.COP mailing list We have two views on this. Strict and Loose. Loose = the /24(ipv4) and /48(ipv6). Strict = using the RIR minimal allocation as the limit. Both work.
signature.asc
Description: Message signed with OpenPGP using GPGMail
_______________________________________________ BCOP mailing list BCOP@nanog.org http://mailman.nanog.org/mailman/listinfo/bcop