WG status, chairs, 10 minutes;
Andrew Alston: early review required for the draft before it is sent
to area review.
Greg Mirsky: BIER ping in the LC but it has not been updated, needs
to talk to the pen holder for the pen switch consideration to move
the draft forward.
Jeffrey Zhang: BIER prefix redistribute draft needs shepherd too.
draft-ietf-bier-mldp-signaling-over-bier, Hooman Bidgoli, 10 minutes
Chairs: will take to the list for LC.
[poll] 16 yes and 0 no for the LC.
draft-ietf-bier-idr-extensions, Jeffrey Zhang, 10 minutes
Chairs: will start in the IDR WG and looking for new shepherd.
[poll] 15 yes and 0 no for the LC.
draft-ietf-bier-prefix-redistribute, Jeffrey Zhang, 5 minutes
Chairs: will start LC and looking for the shepherd.
[poll] 19 yes and 0 no for the LC.
Efficient P4-Based BIER Implementation on Tofino, Steffen Lindner,
15 minutes
Tony P: why there are overlap ports.
Steffen L: for reducing the recirculation. It's an optimization.
Jeffrey Z: the cluster seems like the line card, some
implementations may not need so many iterations.
Gyan M: P4 has the advantage for pipeline programming, are you able
to eliminate the pipelining for recirculation?
Steffen L: just copy the packet to the port which may be forwarded
to. However, this way the egress is blocked by unused packets.
Worst-Case: whole switch is blocked.
Tony P: when you did the static cross group calculation, it may be
low. can the multicast address be configured twice?
Steffen L: it may be not actually multicast address. it's the
identifier which is used to across all the clusters.
Gregory Shepherd: So Tofino has a built in way to process multicast
addresses and map these to ports? (something similar - about the
wording multicast group)
Steffen L: It's an identifier, it's just called multicast group
within the Tofino, it is not a signalled multicast group.
Tony P: too many port may consume IDs.
Steffen L: Yes, e.g., its not possible to only build 2 configured
port clusters with 16 ports each, as we do not have enough multicast
groups on Tofino for that. Therefore, we need to select the
configured port clusters in a clever way.
draft-eckert-bier-cgm2-rbs, Toerless Eckert, 10 minutes
Toerless: AF can be many bits.
Greg S: MTU is the smaller problem.
Jeffrey Z: RBS is better than BIER-TE. like the local BP changing.
Gyan M: if BIER architecture needs to be changed?
Tony P: the bitmask dealing is totally different, but it doesn't
violate the BIER architecture. more tests may be needed.
Ice: The BitMask will likely increase in order to deal with better
encoding.
Ice: There is some data in the control-plane, while it probably
needs to be in the packet.
draft-ietf-bier-frr, Huaimo Chen/Steffen Lindner, 10 minutes
Hooman: if you have basic unicast based LFA, BIER can just follow
it.
Tony P: we may need not always follow IGP LFA.
draft-zwx-bier-te-extensions
no discussion because time expired.