Hi Aijun, > But normally, we need only one router within IGP run BGP-LS.
True - if we keep adding opaque stuff to the IGP to limit the number of BGP-LS sessions per area. In my view BGP-LS should and perhaps will be replaced with a real pub-sub mechanism sooner or later and that will be part of every edge node. Hence trashing the IGP protocol is IMHO a wrong thing to do if the goal here is just to export the information to the controller in the first place. Best, R. On Thu, Jan 13, 2022 at 12:02 PM Aijun Wang <wangai...@tsinghua.org.cn> wrote: > Hi, Robert: > > I remembered we have discussed this. RFC9086 requires every border router > run BGP-LS. > But normally, we need only one router within IGP run BGP-LS. > > I think Tony has gotten one of key use use case for this draft. The > difference between us is how to accomplish it. > > Aijun Wang > China Telecom > > On Jan 13, 2022, at 18:40, Robert Raszuk <rob...@raszuk.net> wrote: > > > Hi Tony, > > If you originate BGP-LS on the PE of interest it seems you can stuff it > with whatever you like. I read RFC9086 as one example of it. > > Thx, > R. > > On Thu, Jan 13, 2022 at 3:05 AM Tony Li <tony...@tony.li> wrote: > >> >> Robert, >> >> On Jan 12, 2022, at 5:06 PM, Robert Raszuk <rob...@raszuk.net> wrote: >> >> Well if that would be controller based TE computation it seems that these >> days BGP-LS (ugh!) would be used instead of IGP flooding to pass that info >> around. >> >> Hence that makes (at least :) two of us pretty puzzled on the real use >> case here. >> >> >> >> BGP-LS can’t pick it up unless it’s in the LSDB. Thus, you inject it >> into the LSDB and let BGP-LS convey it to the controller for you. >> >> Tony >> >>
_______________________________________________ Lsr mailing list Lsr@ietf.org https://www.ietf.org/mailman/listinfo/lsr
