> On Aug 6, 2015, at 17:42, Juliusz Chroboczek <j...@pps.univ-paris-diderot.fr> > wrote: > >> I wasn't aware of the treatment of multicast packets as less than best >> effort in wireless transmission. That is not exactly intuitive, given >> that radio is inherently broadcast. > > Yes, that's counter-intuitive, but actually quite natural. > > 802.11 uses two different MAC sublayers: for unicast, it uses ARQ > (typically 8-persistent) and varies the PHY rate (depending on > e.g. pre-ARQ packet loss), while for multicast, it uses no ARQ and a fixed > PHY rate. > > The effect is that unicast uses a higher data rate and virtually zero > packet loss, while multicast uses a low data rate and suffers from > significant packet loss (20% - 40% is not unusual for full-size frames, > and I've seen 80% on a link that was quite usable, post-ARQ, for reading > mail over IMAP). This is good enough for IPv6 (both ND and RA use little > throughput and deal gracefully with reasonable packet loss), but makes > multicast useless for some other applications.
An additional complication with 802.11 is that various physical encodings use spacial beam forming for unicast and that’s not possible with multicast. It’s the main reason that transmission bit rates for unicast can be so much higher than for multicast— perhaps surprisingly so for people only accustomed to wired physical networking scenarios. It also has murky effects on coexistence between overlapping service sets, which are extremely common in high-density residential environments. —james _______________________________________________ homenet mailing list homenet@ietf.org https://www.ietf.org/mailman/listinfo/homenet
