Re: about draft-ietf-mobike-protocol-01.txt (issue 40)
Jari Arkko <[email protected]>
| Newsgroups | gmane.ietf.mobike |
|---|---|
| Message-ID | <[email protected]> |
>=> what is the real difference between MOBIKE multihoming support >and IKEv2 multihoming support, and, more important, is this difference >enough? > > My understanding of this difference is that in our current proposal, the peers can register their alternative addresses and failover to the other addresses if there's a connectivity problem in the currently used address. (I don't think you can do this in basic IKEv2, and personally, I think its a good tradeoff in terms of functionality vs. effort and configuration.) This is what I call the "failover" multihoming model. In the SCTP multihoming model we can choose more independently where to send data, even if there's still the concept of a primary (default) path. In the load balancing multihoming model we can send even the same stream over multiple paths, and still perform appropriate load balancing, congestion avoidance, and acknowledgement tasks. --Jari