(Didn't have time to add this part before I headed for the outdoors, now that I'm back...)
With a similar set of assumptions, you can also build this: First, in this model we accept that the end stations must, rather than running standard client 802.11b software, take on some of the required intelligence. Second, we assume that we can build a layer 2 mesh network and are comfortable with the side effects and challenges thereof (control of broadcast domains, security aspects of assuming users aren't trying to steal other users' MAC addresses, etc) So, we build that layer 2 network, which connects the APs, both those with local Internet connectivity and not. We then add software to the APs which have local Internet access which: implements a tunnel protocol that can be reached by the end user stations (eg., PPPoE) and which broadcasts to the layer 2 network the availability of that tunnel service. Client stations then come up on the network, listen for what they believe is the "best" access "portal" to the outside Internet, and connect to it. Disadvantages of this over my previous email: Requires special client software Layer 2 network Harder to implement "on-network" services like local (behind the NAT) DNS Advantages: Well understood technology (assuming you adapt PPPoE) Less software to write As always, comments, questions, flames: email me Matthew Kaufman ps. Neither of these suggestions accounts for what I believe to be a serious hole in the *business model* side of wireless today, but I'll save that for a later message -- general wireless list, a bawug thing <http://www.bawug.org/> [un]subscribe: http://lists.bawug.org/mailman/listinfo/wireless
