Barring some kind of mishap, the IEEE was expected to approve the 802.11g standard on June 12 based on version 8.2 of the draft. Of course, products conforming to different 11g draft specs have been available for more than six months. As expected, major vendors will release software upgrades to ensure their hardware complies with the final standard.
Of course, 802.11b isn't going away anytime soon: Its large installed base is increasing every day because of dirt-cheap component pricing and a powerful brand name--Wi-Fi. And let's not forget the lesson we learned when Ethernet beat Token Ring. Like Ethernet, 11b is good enough for most apps.
Managing a hybrid 11b-11g network won't be easy. The fundamental problem is that a legacy 11b device sees an 11g transmission as noise--the same way it sees a poorly designed 2.4-GHz cordless phone. Protection, which takes advantage of a little-used 802.11 feature known as RTS/CTS, solves this problem, but at a significant cost. Assessing that cost has been tricky, especially when dealing with multiple products conforming to multiple draft specs and using chipsets from multiple vendors. Now the situation is becoming a little clearer, though the picture isn't pretty.
First, the good news: In pure 802.11g office-cubicle environments, you'll see real-world TCP throughput of almost 20 Mbps at distances 100 feet or less from an access point. At a range of up to 200 feet, you'll still see throughput greater than 8 Mbps. In both cases, this is almost four times what you see with 11b.
However, in worst-case scenarios, all nodes slow down to the throughput of the slowest device, which may be an 11b device at the fringe of your coverage area, sputtering along at less than 1 Mbps. This should make it clear that 11g is not an effective means of enhancing performance within environments that have a large installed base of 11b devices.