WiFi may be the single most successful internet access technology. It is used by over a billion people. Unregulated use has enabled an explosion of products and deployments using WiFi. Individuals can take immediate action as no network operator has to be asked to install or extend a WiFi network, and this contrasts strongly with centrally managed and deployed systems such as the cellular telephone based communications systems. One WiFi hop is between any company and a large fraction of its users; yet we have paid scant attention to how well WiFi functions, and nearly none at all at how it will continue to scale to the next 300 million hotspots, and 10 billion new users and devices in the next 4 years. WiFi devices now cost as little as US$3, Linux WiFi devices as little as US$9. A large fraction of these devices run/will run Linux, and the current Linux WiFi stack and drivers are far from optimal.
There has been little cross-fertilization between the participants of the IETF, who understand how the Internet’s end to end protocols function, and the IEEE participants in the 802.11 standards process who are primarily radio and hardware engineers. WiFi downward compatibility constraints causes increasing complexity and problems with every succeeding generation of the technology. We must attack the problems in today’s WiFi as it is between us and almost all devices, much or all of the time.
We are focusing on reducing latency throughout the wifi stack, firmware, and hardware. We believe that we can achieve the same sort of performance improvements (an order of magnitude or more) that the Bufferbloat project has already seen in Ethernet based systems.
See the latest presentation and status of the make-wifi-project here
The current working draft is at: https://docs.google.com/document/d/1Se36svYE1Uzpppe1HWnEyat_sAGghB3kE285LElJBW4/edit