100 ft might be a more typical overlap. I'd consider using adhoc mode and just sending the packets out from the mobile client and letting each fixed point forward anything it hears (ie, multipath routing). On the reverse path (fixed to mobile), send the packets to both of the fixed nodes and let them both send it out to the mobile node. Position prediction should work well on a highway (ie, it is easy to predict what the next fixed node will be).
Both directions require throwing away any duplicate packets, but this is easy. As much as possible, you want a stateless system. I'd use packet encryption as the authentication model. Quoting Bernard Aboba <[EMAIL PROTECTED]>: > > We are engaged in a project that needs to provide internet access in a > > highway 150 kms with seamless mobility. > > You should understand that this is going to be a challenging project. At > vehicular velocities, the amount of time spent in the coverage area, let > alone the coverage overlap area, isn't very large. For example, 60 mph = > 88 feet/second. If the coverage overlap area is 10 feet, this means that > the vehicle is only present in that area for 114 ms. In that time it is > necessary to complete the 802.11 scan, authentication and association as > well IP layer network attachment detection, in order for connectivity to > not be interrupted. > > Bill Arbaugh's team at the University of Maryland has been doing > measurements and as I understand it, scan times alone can run 40-400 ms. > With WPA, the 4-way handshake is always run, and depending on the client > and AP, this can take as little as 10 ms or more than 40 ms. > > In a highway situation, key caching won't help much because vehicles will > typically not retrace their steps within a time period in which the cached > keys would remain valid. > > Therefore, it probably makes sense to do either pre-authentication or > pre-emptive key generation. With pre-authentication, it is probably > necessary for the AP to advertise the neighbor graph in the Beacon in > order to allow pre-authentication to begin prior to entering the coverage > overlap area. Otherwise pre-authentication would be unlikely to complete > in time; measurements show authentication times from 250 ms (fast resume) > to 700+ ms (full EAP TLS authentication). > > For references to papers on pre-authentication, pre-emptive key > distribution and other aspects of 802.11 handoff, see: > > http://www.drizzle.com/~aboba/IEEE/ > -- > general wireless list, a bawug thing <http://www.bawug.org/> > [un]subscribe: http://lists.bawug.org/mailman/listinfo/wireless > > -- general wireless list, a bawug thing <http://www.bawug.org/> [un]subscribe: http://lists.bawug.org/mailman/listinfo/wireless
