> > I think that the Point-to-Point idea has some merits, > but the applicability is not general. > > I think that if we're going to keep the ethernet-like > nature of WLAN (it is a deployed base) then the > hosts which are aware of their performance issues > need to respond to them. > > I'll see if I can come up with a reasoned description > of my idea in a draft. > > (reasoned may be the difficult part for me). > It's not necessarily about deprecating ethernet emulation but maybe about adding additional possible abstraction for L3 to use over the same MAC. This could be negotiated on a per association basis, defaulting to ethernet.
Anyway the initial problem stands, even if the solution that came with it is debatable. There needs to be an adaptation to the Neighbour Discovery mechanism when used over WLAN because the nature of the network makes it inefficient. Even when using ethernet emulation, there is still some knowledge somewhere that there is a hub and spoke behind. I believe that the proposal must use that and give a specific role to the AP in order to avoid the use of multicast. I'm not sure it means piggybacking with the beacons at all. But I think it's asymmetrical. An AP based range as I succinctly described has some merits, but then we have to look at the interaction with movements. Like if you get your address from AP1 and then move to AP2, AP2 may want to check with AP1 that the MAC/IP address is a correct match. Pascal -------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
