On Mon, Jul 15, 2013 at 11:27 PM, Antonio Querubin tony@lavanauts.org wrote:
(Please trim inclusions from previous messages) _______________________________________________ On Mon, 15 Jul 2013, Heikki Hannikainen wrote:
But but... I think they absolutely must stay in encap.txt even if a BGP announcement is place!
Existance of a route in the encap file implies there is a tunnel established at the other end willing to accept the encapsulated traffic. The sites doing BGP may or may not be doing that. If the latter, then you're just sending traffic to a black hole.
I would suggest that all of the BGP-enabled sites should have tunnel endpoints like before, so that all the old non-BGP net-44 sites can talk directly to the BGP sites, even when they old sites are behind uRPF/spoofing filters and cannot transmit unencapsulated net-44 traffic.
If one of the major reasons to do BGP announcements in the first place is to reduce traffic going via UCSD, then it makes a lot of sense to keep the old tunnel endpoints in place, and not have all traffic from non-BGP sites to BGP sites go via UCSD. It's good for latency and resiliency.
- Hessu