> Jon Crowcroft wrote: > > > i suggest reading > https://www.icir.org/floyd/srm.html > > > I did so before my previous reply. > > But, it did not affect my understanding that there are so > many complicated protocols for reliable multicast. As such, > TCP mesh should be better as long as we follow the CATENET > model. > > > plus any of many papers on tcp incast > > > I'm afraid that if TCP incast is a problem for control traffic, > there should be no bandwidth left for data traffic.
incast isn't about persistent congestion, it is about synchronised sources. > > If lack of bandwidth is a problem for data traffic, the way to > go is QoS routing, which means, over QoS satisfied routes, we > have enough bandwidth for control traffic. > > Masataka Ohta > >
