On 9/Sep/20 09:21, Robert Raszuk wrote:
Nope .. it is the other way around.
It is all easy if you look from your network centric view.
But if I am connected to 10 ISPs in each POP I have to build 10 different egress policies, each embedding custom policy, teach NOC to understand it etc...
Well, yes and no. We, for example, connect to more than one transit provider in at least one of our PoP's. The outbound policies are exactly the same. The only difference is the differences in naming for each policy, and how we signal RTBH into the transit network. Everything else is the same. Rinse an repeat for all the exchange points we have connected to a single router, in a single PoP. I get that no two BGP routing policies are the same between operators, but I'm not certain standardizing on communities is going to make things any less complicated than we currently assume they are.
I think if there is a defined way to express prepend N times to my ISP peers across all uplinks or lower local pref in my ISP network in a same way to group of ISPs I see the value.
These kinds of policies are generally do-and-forget. When you spend time turning up a new provider, you are dedicating time to setting them up on your end. An extra 3 minutes to configure communities they have published is not overly complex, I believe. Moreover, it's not something you are likely to revisit outside of a communicated change on their end, or troubleshooting on your end. I'm all for making many things as standard as possible, but if our goal is to make signaling of external communities simpler than it is today, I don't quite see how standardizing said communities will simplify that process in a practical sense, on a global basis. As always, I could be wrong... Mark.