Chair -------- waiting for LC on large comms - though accidentally submitted add paths in editor Q gshut likely reopened openbmp presentation ------ Randy/Serpil present slides - Pleas from the developers for data sources. - Job comments that tagging routes to route collectors with communities might be bad use of comms ruediger comments; but i couldnt hear it. randy says that it could just be a email from the source randy cuts it off; saying for later discussion along with draft in preso.R - Plea to come to bar BoF right after WG mtg Loc-rib/lov-rib-out preso ------- Serpil presents - bgp-ls is a big driver for this local-rib export - randy bush suggestes that theres no real rib-out evans says its a representation - raszuk suggest u want loc rib and loc-fib to have whole picture. or maybe one or other Evans says they're concentrating on control plane. maybe future. - Scudder asks about state compression; suggests please DONT say NOT to do state compression. its puts backpressure on the control plane and bound to fail. Evans says they aim to not make the collector a bottleneck. but some esp. those doing bpg-ls want ALL events and not lose anything. a lot can happen in the 1s compression window. Scudder can't understand 1s - where does that come from? *** needs more disucssion Serpil says they want as much as possible - Egor suggest mights be difficult with all possible ECMP routers. but missing recursion view and you dont know that its going to do ECMP. Evans says they look at bgp NHs and with BGP-ls can look down that hole. Egor - hard to maintain compatibility with Impl.s. i missed continuation of this comment from igor - Chair asks if folks want to adopt this draft. overwhelmingly yes. - Randy - suggests is protocol so might be IDR. PDS says he'll chairs BGP sesssion culling ------ Job presents & talks excessively as usual ;) - jen linkova - suggests route filters instead of shutdowns Job says there severals ways to drain traffic. but trying to capture that there are L1 or other player sthat may not make it obvious. Jen please mention filters and other ways to do this and notify cusytomer - colin RIPE - mention BGP GR, b/c traffic will keep fwd'g Job says there can be statics too. its a call on traffic levels colin - maybe point that out [sic] - warren kumari - its horrifying that we'd need this. but... also think that the terminology used in the draft is odd and needs work Job - says there are more than one point of view so need differentiate - Ruediger DT - as user of the fabric, why not use BFD? Job - BFD may not always be available Joel - BFD may not tell you what you want to know Ruediger - things could cause blocking of BFD or the other mechanisms - Joel Fastly - doesnt work well on multi-lateral XPs with multiple Switches and the maint isnt on your switch. lacks note that this is permanent state one some routers (or 2 config changes) but just temporary state on others - or can be so. - Raszuk - like tcp fwd blocking. what about VXLAN, and other L2.5 services. Job thinks it might be out of scope. but conceptually, disrupting the routing is the goal - to drain the traffic forwarding. - Job asked chair for WG adoption. what do folks think? no one speaks. Chair - will take WG adoption to GROW ML.