I'm wondering... what if, in the hypothetical routing protocol "Babble", one got rid of multicast hellos and ETX entirely, and routed using RTT measured by unicast hellos. Wouldn't this take delays from ARQ'ed packet loss into account have results somehow similar to ETX?
On Tue, Jun 20, 2017 at 3:06 PM, Juliusz Chroboczek <j...@irif.fr> wrote: > >> 2. 802.11 uses adaptative rate control on unicast frames, but not on > >> multicast frames. > > > This is not entirely true, since hostapd is able to send multicast as > > unicast-frames. > > And this will cause Babel's link-quality estimation to yield incorrect > values. > > Dynamically computed metrics are an open area of research, and one that is > sadly neglected by the scientific community. I believe that it is > important, but it is very difficult to publish in that area (FWIW, our > paper about the RTT metric was rejected twice). > > ETX is an elegant hack that uses the fact that 802.11 multicast isn't > subject to ARQ. As you justly note, Ruben, this assumption can fail in > edge cases, and there's nothing that can be done about it. The obvious > solution would be to have explicit cross-layer signalling, as in the case > of Lamparter's variant of IS-IS. Doing that right is not easy, and might > (or might not) require changes to the link-layer. > > I'm personally interested in that area of research, and I'll probably have > a fair amount of funding in the near future. However, since the area is > not fashionable, I cannot promise that we'll be able to publish. > > -- Juliusz > > _______________________________________________ > Babel-users mailing list > Babel-users@lists.alioth.debian.org > http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/babel-users >
_______________________________________________ Babel-users mailing list Babel-users@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/babel-users
