Failover only works for clients. For server p2p you do not have such a fallback and servers often have a private and a public IP.
Kind Regards, Mikkel Fahnøe Jørgensen On 14 October 2020 at 17.49.05, Martin Duke ([email protected]) wrote: Gorry, that is my understanding as well. If we don't have failover, we've messed up in some way. On Wed, Oct 14, 2020 at 3:23 AM Gorry Fairhurst <[email protected]> wrote: > Maybe it's just me - but I would be super-interesting in checking that I > really understand what the migration feature of QUIC would give us if > there was a change in connectivity. > > Am I correct, in seeing QUIC as already having (or nearly having) a path > failover method - similar to SCTP failover? ... where multiple paths are > kept alive, but only one path is used at a time for data. Can some > problems be solved by a simple failover method? > > To me multipath, is where performance optimisation questions are solved > - such as how to have simultaneous use of capacity across multiple paths > (some people call this bonding), or scheduling across paths to > minimising the delay, or one of many other metrics. > > Just a thought on what might be useful to me to see in some slides .... > perhaps before we delive into the details of policies and scheduling. > > Gorry > > > On 14/10/2020 10:45, Lars Eggert wrote: > > Hi, > > > > On 2020-10-13, at 18:35, Spencer Dawkins at IETF < > [email protected]> wrote: > >> I'd like to do one or two slides, five minutes max, on the multipath > modes that are under discussion in 3GPP SA2, from my perspective as an > individual. > > I think that would be helpful, and so far there are no other offers to > summarize those use case requirements. > > > >> When do you need the slides? > > Monday at the latest? > > > > Thanks, > > Lars > > >
